Overall, cancer is the second leading cause of death in the United States, exceeded only by heart disease.⁸ It is estimated that within the next 10 to 15 years cancer will overtake cardiovascular disease to become the most common cause of death in America.^9,10 (If cancer were a disease with no associated mortality and evenly distributed throughout the population, and, assuming a lifetime cumulative incidence of approximately 33%, one would expect that 1 in 9 people would develop two primary cancers in his or her lifetime. Taking this a step further, 1 in 27 would develop three primary cancers, and so on). Chance or random distribution probably plays the most important role in SMNs; however, a significantly increased frequency of association between two tumor types often occurs and may point to a specific etiology. Sections on genetic risk factors and therapy-related secondary cancers discuss specific etiologies. However, while those may seem most dramatic, shared risk factors probably play the greatest role in the development of SMN.

Analyses from the US National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) Program in 1995 indicate that multiple primary neoplasms comprised 13.1% of cancers in men and 13.7% of cancers in women, a proportion that had doubled over the previous 20 years.^1,2 This refers to the proportion of secondary cancers among all cancers, but when one compares cancer survivors with the general population, the incidence rate of SMNs appears to be almost twice the rate of initial primary cancers. Stated in a different way, any survivor of cancer has approximately twice the probability of developing a new primary cancer than a cancer-free individual of the same age and sex.^1,11

A critical factor determining the increased risk of developing a SMN is the probability of surviving a first neoplasm. The prevalence of SMNs in childhood cancer survivors, comprising more than 13,000 individuals diagnosed between 1970 and 1986 (the Childhood Cancer Survivor Study), was recently published.¹² The cumulative incidence within 20 years from diagnosis was 3.2%. When adjusted for radiation therapy, the highest risks were associated with younger age at diagnosis, a diagnosis of soft tissue sarcoma or Hodgkin's disease, and treatment with alkylating agents.

Another potential factor is detection bias. During the work-up of a patient with an initial symptomatic cancer, otherwise clinically unrecognized malignancies may be detected. Thus, most experts distinguish between malignancies discovered or diagnosed simultaneously and those that are metachronous or develop subsequently. Detection bias is also seen with the enhanced long-term surveillance for SMNs in cancer survivors.

Although case reports and case series describe certain pairs or clusters of cancers in individuals, the establishment of a true pair-wise association requires a more systematic and controlled approach. Cancer may be common, but individual types of cancer are relatively rare. For example, breast cancer and prostate cancer occur in 1 to 2 per 1,000 individuals per year at the ages associated with the highest risk for these cancers. Thus, a large number of cancer survivors are needed with one type of initial cancer in order to determine whether the development of certain SMNs are a result of chance or are true associations.¹³

As a result, the most valuable information concerning the incidence of SMNs has come from population-based registries rather than hospital-based series. In addition to providing follow-up on large numbers of cancer survivors, cancer registries also provide sufficient information to calculate the expected rates of SMNs. A comparison of the observed incidence with expected incidence provides the standardized incidence ratio (SIR), an estimate of the risk of developing SMN in a given survivor. For example, a SIR of 2.0 would indicate that an individual diagnosed with a given malignancy is at twice the risk of developing a second malignancy of a given type, compared with a similarly aged member of the general population of the same sex. Such rates and ratios generally exclude the initial 3 to 6 months after diagnosis in order to eliminate synchronous malignancies. Furthermore, estimates are made for different time intervals following the initial cancer in order to evaluate the effects of enhanced surveillance on the observed incidence of SMN.^11,13–16

The most important contribution in this area, to date, is a systematic analysis of all major cancers as second malignancies, published as a National Cancer Institute monograph in 1985.¹⁴ This analysis used data from the Connecticut Tumor Registry, the oldest extant systematic population-based registry available, and similar data from the Danish Cancer Registry, another long-ongoing population-based cancer registry. More recently, multiple studies have used data from the SEER program, which comprises population-based registries encompassing approximately 10% to 14% of the United States population. The SEER program had amassed data on 1.8 million new cancer cases diagnosed from 1973 to 1992. The interested reader is referred to ref. 15 for a systematic analysis of pair-wise risk for selected cancers using this data set. Table 159-1 includes data from the analyses of SEER data for illustrative and discussion purposes.

Table 159-1

Risk of Secondary Cancers for Selected Cancers, Based on Data from SEER, 1973-1994.

An excellent illustration of the bias that occurs in the detection of SMNs from enhanced use of diagnostic tools at the time of initial work-up or for subsequent surveillance of a cancer patient is the interrelationship between colorectal cancer and prostate cancer in men. Studies show that following the diagnosis of colorectal cancer in men, there is an enhanced diagnosis of prostate cancer secondary to increased screening for prostate cancer. However, men diagnosed with prostate cancer do not have an elevated incidence of colorectal cancer, presumably because there is not as much clinically occult colorectal cancer waiting to be diagnosed under these circumstances.^{14,15,17–19}

The above illustrates an important principle, that is, one can use the directionality of the pair-wise association as a means for investigating the etiology of the association. For example, a therapy-related SMN should occur at increased incidence following a specific initial primary cancer, such as leukemia following Hodgkin's disease. The association should be, and in fact is, unidirectional, that is, there is no enhanced risk of Hodgkin's disease following leukemia.^14,15,17 When shared environmental exposures play a major role in the elevation of risk, one expects to see a bi-directional association, that is, each member of the pair should occur at elevated risk following the other primary malignancy.

Another illustration of this phenomenon is bilateral disease, or development of the same malignancy in the second of paired organs. For almost all cancers, there is an elevated risk of the malignancy in both paired organs, for example, bilateral breast cancer or bilateral ovarian cancer. Whether these are diagnosed synchronously or metachronously, it seems likely that this is a result of shared factors, whether environmental or genetic. Other portions of the same organ may also be affected, as in multiple primary cancers of the colon.^14,20,21