The Cancer Journal - Volume 10, Number 2 (March-April 1997)
Watch your prostate!
Medicine in the grip of technology - Although health is the main concern of medicine, its exact meaning is not at all clear. Medicine is more concerned with non-health, known as disease. Yet not every disease should be regarded as non-health. Many young adults carry in their arteries arteriosclerotic plaques without any apparent health deterioration. Are these plaques non-health and should they be treated? Obviously health and disease are not complementary. How then should diseases that do not impinge upon health be handled? Sophisticated technology reveals slight aberrations that in the past were unknown, and medicine is confronted with a new problem, how to deal with aberrations in a patient that feels healthy? Medical confusion is most pronounced in the treatment of breast cancer (1), and also affects patients with prostate cancer.
Rising cancer prevalence with age - Cancer prevalence rises with age, and in some cancers, like that of the prostate, it rises exponentially. In autopsies of men who died of causes other than cancer, carcinoma of prostate was found unexpectedly in 26-37% of prostates (2). Such cancers were called incidental. Over 40% of men in their sixties with normal rectal examinations had histological evidence of prostate cancer (3). The increase in prevalence may indicate that in old age, cancer is less aggressive, and grows more slowly (4). Incidental cancers might be slow-growing aberrations that under normal circumstances would never be detected. Yet rising cancer prevalence is also documented by epidemiological studies, suggesting that proneness to cancer is rising. Closer analysis of these studies reveals that most of them have fallen prey to a serious sampling bias. As diagnostic techniques become refined they reveal smaller and smaller aberrations whose biology differs from clinical cancers. These findings constitute the bulk of the recently detected cancers which contribute to the rising prevalence (5). More and more incidental cancers are included in epidemiological reports. While prevalence rises, cancer biology does not change.
A simple illustration - The figure 1 illustrates hypothetical tumor growth curves of cancers in six individuals. The left curve starts when a single cell is transformed in to a tumor which starts growing. The curve crosses five horizontal lines, or detection thresholds. When it crosses the first, the tumor was detected by genetic means. The second marks the event when the tumor was detected by chance, as an incidental cancer. The third line indicates that the tumor was detected by a biopsy. The fourth line marks the event when the tumor was detected by rectal examination. When the curve hits the upper line, the patient dies from prostate cancer. The figure depicts the fate of untreated tumors. Each of the horizontal lines represents a detection threshold through which the tumor surfaces. The lower the threshold, the more sensitive the detection method. The upper line stands for the event in which the disease was diagnosed only after death.
The second curve crosses only four horizontal lines, and does not reach the upper one. Instead it crosses the vertical line that stands for death from other causes. When a patient dies from myocardial infarction he carries to his grave a small tumor that in itself did not kill him. Suppose that prostate cancer was detected solely by rectal examination. Since the two curves cross, or surface through, this detection threshold, rectal examination revealed only two tumors in this example. Tumor surfacing marks the beginning of the clinical phase of cancer. The number of growth curves crossing the detection threshold, here two, is called cancer incidence, and the number of curves above the threshold, also two, is called cancer prevalence. The other four curves below the rectal examination threshold were undetected.
Lead-time bias - Now suppose that rectal examination was replaced by routine biopsy of all males. Since this method detects smaller tumors, the threshold is lower. Tumor incidence, and prevalence are three. Cancer mortality remains one. Two patients died from other causes. By improving the diagnostic method prevalence rises, yet tumor biology remains the same. The tumor is detected 5 years earlier. However, its biology did not change, and the patient survives with his tumor five years longer. This is known as lead-time bias. Suppose that after biopsy was imposed upon the population, the patients also received a new cancer drug, the improved survival would be attributed to the drug. Although epidemiologists are aware of this bias, it is seldom corrected. In order to avoid this bias, the new treatment ought to be related only to the upper detection threshold. In other words, only patents who were diagnosed by rectal examination should participate in the clinical trial that evaluates the efficacy of the new drug. All other patients ought to be left out. Unfortunately, this requirement is hardly ever met.
Length bias - The figure illustrates that with age, the curves are less steep and tumor growth is slower. Prostate cancer detected incidentally is represented by a lower horizontal line, which is crossed by five curves. Cancer incidence, and prevalence are now five. Cancer mortality remains one. As diagnostic means become more and more refined, more slowly growing tumors are detected. Some require decades to reach the upper cancer death line, and are relatively harmless, yet they are included in the statistics. This is known as length bias. Genetic methods reveal six tumors. Prevalence and incidence are now 6, more tumors were detected at an earlier stage, yet the overall tumor biology did not change. Since genetic methods do not distinguish between slow- and fast-growing tumors, the mere detection of a cancer gene is meaningless. In other words, a young female with a BRCA1 gene may carry a slow growing tumor without any clinical significance. Should she remove her breasts as some genetic councilors would suggest? This illustrates the real threat to individuals who participate in genetic screening (1).
This absurd situation is caused by two conceptual flaws: 1. A complex process such as cancer, is represented by epidemiology by only two points in time: incidence and mortality. They do not suffice to distinguish between dangerous and non-dangerous tumors.. Take for instance incidental tumors that are less dangerous to the patient than overt tumors. The two tumor types cannot be equated, or pooled together. Each tumor type poses a certain threat to the patient, and this important information is not accounted for by epidemiology. It is like pooling apples and oranges only because they are round. 2. Cancer definition is incomplete. Since cancer threat is related to the method by which it is detected, this information should be included in the term cancer. The prevalence statistics ought to specify the diagnostic method that detected a cancer, otherwise they are unreliable.
The Swedish example - In a thorough study, Swedish epidemiologists demonstrated that these two biases were responsible for the rise in prostate cancer prevalence in Sweden (6). "The incidence of prostate cancer has increased during the past 30 years but has been paralleled by increases in survival rates from this disease, despite the absence of documented major improvement in curative treatment." They analyzed a population-based cohort comprising all cases of prostate cancer (n=80,901) detected in Sweden during the period of 1960-1988. "Most of the major improvement in survival rates is probably not a result of improved treatment but increased detection of non lethal tumors and lead-time bias." The mere detection of a tumor does not reveal its biology, which remains the missing dimension of cancer. Under these circumstance the only sound policy for treating localized cancer of the prostate should be watchful waiting or palliative hormonal therapy. This is how most Swedish physicians treated their patients (6).
e-mail: Gershom@md2.huji.ac.il Internet home page: http://www.md.huji.ac.il/special/cancer/journal.html
1. Zajicek G. Lady, think twice before having a mammography. The Cancer J 9, 172-173, 1996.
2. Anderson's Pathology.( Kissane JM Ed); C.V. Mossby Co, St. Louis, p 907, 1990.
3. Montie JE, Wood DP Jr, Pontes E et al. Adenocarcinoma of the prostate in cytoprostatectomy specimens removed for bladder cancer. Cancer 63, 381-385,1989.
4. Zajicek G. Cancer in old age is more benign than in younger adults. The Cancer J 9, 64-65, 1996.
5. Black WC, Welch HG. Advances in diagnostic imaging and overestimation of disease prevalence and the benefits of therapy. NEJM 328, 1237-1243,1993.
6. Helgesen F, Holmberg L, Johansson JE et al. Trends in prostate cancer survival in Sweden, 1960 through 1988: evidence of increasing diagnosis of non lethal tumors. J Nat Canc Inst 88, 1216-1221, 1996.