Several steps are needed to turn a normal cell into a cancer cell. Most, if not all, of these involve mutational change. Cancer is therefore a genetic disease at the level of the cell. Cancer may also be in some cases a genetic disease at the level of inheritance. This chapter will consider the inherited contribution to cancer incidence, its importance and its investigation.

Inherited predisposition to cancer is potentially important for three reasons. First, individuals who are predisposed to cancer are a high-risk population who may benefit from efforts at prevention, or from early diagnosis and treatment. Second, genetic predisposition may tell us about the development of cancers in general. Most, if not all, cancers that have an inherited predisposition have a similar non-heritable counterpart; in the cases so far studied, heritable and non-heritable cases are genetically similar. Third, many of the genes which, when mutated, are ‘cancer genes’, may have important normal functions in growth and development. Cancer families are a way in which these genes are revealed and can be identified.

There is no clear answer to this question. Each case will result from a mixture of inherited predisposition, environmental exposure and luck, since we are dealing with the accumulation of a series of chance events. The question can be rephrased: which cancers have an inherited predisposition that is sufficiently strong to be of practical significance for prevention or treatment. The answer to this is also still unclear. Perhaps less than 1 per cent of cancers occur in the ‘inherited cancer syndromes’, described below, and a further 5 to 10 per cent in recognizable familial clusters that probably have a genetic basis. However, inherited predisposition which is too subtle to cause an obvious family history may be more important.

It can be difficult or impossible to determine whether a cancer in an individual is due to inherited predisposition. In a few rare syndromes, there is a ‘marker phenotype’, for example the multiple intestinal polyps characteristic of the inherited cancer syndrome familial polyposis. The clinical and family history may be indicative. If the cancer is rare, this is straightforward: the occurrence of several cases in the family, or of multiple primaries in the individual, is noticeable and almost certainly significant. If the cancer is common, however, the significance of several cases in a family may be much more difficult to determine. This point will be discussed further in relation to breast and ovarian cancer.

The inherited cancer syndromes include all those cancers in which a genetic effect is clearly apparent. Together they probably account for less than 1 per cent of cancer cases.

In some cases (e.g. the polyps in familial adenomatous polyposis) the marker phenotype is the result of a growth abnormality in the ‘target’ tissue from which the cancer will arise, and can be regarded as a preneoplastic lesion. In other cases, such as the multiple cysts of abdominal organs seen in von Hippel-Lindau syndrome, the phenotype results from developmental defects in tissues remote from the cancer, and presumably reflects a pleiotropic effect of the inherited mutation. The diversity of tissues and effects (illustrated in Table 2 665) is puzzling because in most cases it does not conform to current ideas of lineage or physiological relationships.

In familial adenomatous polyposis, as in all of the inherited cancer syndromes so far recognized, the pattern of cancers in the family is that expected of an autosomal dominant predisposing gene; that is, the offspring of an affected person have a 50 per cent chance of inheriting the gene. In a few special cases, such as glomus tumours and Beckwith-Wiedemann syndrome, the expression of the disease is modified according to the sex of the parent who transmitted the gene: this is the phenomenon of ‘genomic imprinting’.

Predisposition is not to cancers in general, but to cancers of specific sites (Table 2) 665. In familial adenomatous polyposis, for example, the cancers which occur are, in addition to colorectal carcinoma, carcinoma of the ampulla of Vater, hepatoblastoma, and thyroid tumours.

Variation may be seen in the spectrum of tumours and other phenotypic abnormalities, or in the probability that each will be manifest by a certain age. This variation may ‘breed true’ in families, in which case it is presumably due to different predisposing mutations (either at different loci, or different mutant alleles at the same locus). Alternatively, it may occur even within a single family, in which case it cannot be due to differences in the predisposing mutation (which is the same in all family members) but must be due to some combination of ‘modifying’ genes, environment, and chance.

Several distinct syndromes are recognized. The extent to which these different clinical syndromes are reflected in genetic differences is still unclear. Familial adenomatous polyposis is defined genetically by mutation of a specific gene located on chromosome 5q21. Predisposition due to mutation of a gene on chromosome 18 has been suggested for ‘non-polyposis’ colorectal cancer families, although this requires to be confirmed. Recently, however, a family in which inherited colonic cancer was associated, on average, with numbers of polyps intermediate between those typically seen in familial adenomatous polyposis and in ‘cancer family syndrome’ was reported also to be due to an inherited mutation of the familial adenomatous polyposis gene on chromosome 5. There may therefore be a spectrum of familial colorectal cancers associated with adenomatous polyps, and with risk of other cancers to varying degree, that are due to different mutant alleles at this locus. A similar situation is seen in the multiple endocrine neoplasia type 2 (MEN 2) syndrome. Three clinically distinct varieties of this syndrome breed true, but map to the same region of chromosome 10, suggesting they may be due to different mutations of the same gene. The classification of different clinical syndromes of inherited cancers will become much clearer as the genes are isolated.

Some members of families with typical familial adenomatous polyposis (colorectal carcinoma, multiple intestinal polyps) also have pancreatic ampullary tumours, desmoid tumours, and mandibular osteomas (Table 2) 665. This was originally delineated as a separate clinical variety, ‘Gardner's syndrome’: it is now clear that the so-called ‘Gardner's phenotype’ occurs to a greater or lesser degree in most kindreds with familial adenomatous polyposis. Gardner's syndrome cannot therefore be the result of different inherited mutations: the range of expression within a family must be the result of modifying influences (as yet undefined) acting on the expression of the same mutation in different family members.

The proportion of carriers of the familial adenomatous polyposis gene who develop colorectal polyps increases with age. Even by age 25, about 10 per cent of carriers do not have detectable polyps. This variation may be due to the effects of other, ‘modifying’, genes, to environmental effects, or to chance: dissection of these factors is difficult. From a clinical standpoint, the probability of a marker phenotype or cancer being present by a given age is important in family screening. Such data are needed to enable estimation of the probability that a family member with a negative screen or with apparently unaffected parents at a given age may still be a gene carrier.

Unfortunately, it is still not possible to predict the penetrance or expression of the predisposing gene in an individual family member.

Some families are at high risk of a common cancer, such as breast, colon, or ovary. Almost every type of cancer has been reported to occur in several members of a single family at some time. Is the clustering significant? If so, is it due to inherited factors or to other factors? The final proof of genetic predisposition comes from demonstration of genetic linkage (that is, the consistent inheritance of a specific genetic marker in family members with the cancer) or the finding of the mutation itself in affected families. Until then, a marker phenotype such as those in the inherited cancer syndromes would provide strong evidence. Since these are absent in the familial cancers, the assessment must be based on the strength and pattern of the familial association in each case.

Solid evidence for familial clustering must come from truly population-based studies, which are free of the bias caused by collecting ‘interesting’ families in a clinic. The usual indicator of familial clustering in such studies is the risk of cancer in the close relatives of an affected individual, compared to the risk in the population in general. The analysis may be stratified to look at groups who might be expected to be at higher risk, such as the relatives of cases diagnosed before the age of 50, or relatives in a family where there are already two individuals affected.

For most common cancers, the increased risk for siblings is of the order of two- to three-fold (Table 4) 667. Although a risk ratio tells us that cancer really does cluster in families, it does not indicate the clustering is of genetic origin, or how the risk is distributed between individuals, because the relative risk is estimated as an average over the whole population. The same overall risk might in principle result either from a common gene which confers a slight increase in risk to many people, or from a rare gene with a very strong effect in a few. These questions can be approached by ‘segregation analysis’, which attempts to find the most likely explanation for an observed familial clustering of cancer by analysis of the pattern of occurrence of the cancer in the families of a large series of cases, ascertained without knowledge of family history.

Almost all such analyses of the common cancers (breast, ovary, colon) agree that the most likely explanation for the observed familial clustering is dominant predisposition by an uncommon gene with a strong effect which affects a small number of families.

The familial cancers resemble the inherited cancer syndromes in many of their features.

Clinical observation of pedigrees, as well as population-based estimates of relative risks to siblings, indicate that specific cancers are associated in familial clusters. Some of these form well-known syndromes, such as site-specific colonic cancer (‘Lynch type I’) and ‘family cancer syndrome’ (‘Lynch type II’), characterized by cancer of the colorectum, uterus, ovary, stomach, and breast. Neither the boundaries between different ‘syndromes’ nor the cancers which belong to each one are, however, always clear. For example, breast and ovarian cancer are often associated, but the clinical spectrum runs from so-called ‘site-specific’ breast cancer, through ‘breast - ovarian’ families with each cancer in different proportions to ‘site-specific’ ovarian cancer ( Fig. 1(a 2774,b,c)). There may be genetically determined differences in age at onset of breast cancer in different families. The Li-Fraumeni (or SBLA—sarcoma, breast, lung, adrenal) syndrome is easy to recognize in its most florid form, which is quite uncommon, but there are many more families in which a woman who develops breast cancer at a young age has a single relative with a sarcoma or a brain tumour: do they fit? Recently, some typical Li-Fraumeni families have been shown to have germline mutations at the p53 locus, and predisposition in some extensive families prone to the development of breast cancer at a young age has been shown to be linked to a locus on chromosome 17q. Findings such as this will be the key to genetic classification of these families. Genetic markers will also resolve the question of which cancers are really part of a given syndrome, because it will be known whether their incidence in known gene carriers is really higher than that in the general population. Similarly, within a single family, markers will indicate which cancers can be attributed to inherited predisposition, and which are merely chance.

So far, we have considered only cases in which inherited predisposition is strong enough to cause obvious family clustering of cancer. However, a slight reduction in the penetrance of the predisposing gene—that is, if only a few gene carriers actually develop cancer—can still cause significant predisposition, even though there are seldom enough close relatives actually affected to cause an obvious family cluster. Such calculations suggest that familial cancers may be only the tip of the iceberg of inherited predisposition, and that in public health terms, the effects of common genes of weaker effect may be much more significant.

For example, a dominant gene that results in cancer at a specific site in only 1 in 10 of gene carriers will hardly cause any extended family clustering, because the risk is only 1/2 × 1/10 = 1/20 in each of the close relatives, and (on average) less in more distant relatives who are less likely to have the gene. Even so, over a range of plausible assumptions about gene frequency and relative risk in gene carriers, such a gene can lead to a remarkable concentration of risk in a predisposed minority of the population. To take a worked example, if the gene confers a 100-fold increase in risk, from 1 in 1000 to 1 in 10, and it has a frequency in the population of 0.1, it will result in an overall incidence of the cancer of 1 in 50. Ninety-five per cent of affected individuals will, however, be drawn from the 19 per cent of the population who have the predisposing gene. (The frequency distribution of alleles is given by the Hardy-Weinberg equilibrium: p&sub2;; 2pq; q&sub2;. If the frequency of the predisposing allele q = 0.1, then the wild-type gene frequency p = 0.9; 81 per cent of the population will be homozygous pp, 18 per cent heterozygous, and 1 per cent homozygous qq.) The relative risk of cancer to siblings of a patient with the cancer (the usual measure of familial clustering) will be about 2.9-fold—roughly the value which is in fact found for many of the common cancers (Table 4) 667. In both ovarian and lung cancer the relative risk to siblings is around three-fold. However, in ovarian cancer much of this risk is probably accounted for by a few families at very high risk, many of whom are recognizable as multiple case families. By contrast lung cancer rarely shows family clusters. Here, there is presumably a more even distribution of risk among the population, caused by a more common gene of less effect.

Recognition of the inherited cancer syndromes and the familial cancers occurs through familial clusters and marker phenotypes. Moreover, the genes involved can be found, with no previous knowledge of what they are, by the methods of genetic linkage, which exploit the occurrence of several cases in a family. In the absence of family clustering or of obvious marker phenotypes, one has to start with ‘candidate’ genes—that is, by guessing which genes may be involved, and testing them in case-control studies of subjects with and without cancer. The candidates which have so far attracted most interest are those whose genetic variation might affect interaction with potential environmental carcinogens. Attention has focused in particular on the genes of the cytochrome P-450 system, which is involved in the metabolism of a variety of compounds, including polycyclic hydrocarbons and endogenous and exogenous steroid hormones. Other enzymes involved in conjugation and detoxification and which are known to exhibit genetic variation, such as acetylating enzymes, have also been studied. There is recent evidence, not yet conclusive, to suggest that variation at the locus for cytochrome P-450 CYP1A1 may determine lung cancer risk in cigarette smokers, and that acetylator status is associated with bladder cancer risk.

Heterozygotes for the gene defects associated with recessively inherited DNA repair syndromes may be another group in this category. Although still controversial, there is increasing evidence from studies of relatives of cases that heterozygotes for the recessive disorder ataxia telangiectasia may be at increased risk of developing several of the common cancers. Depending on the frequency of the ataxia telangiectasia gene in the population, this might account for a significant minority of these cancers in young people.

Precise definition of the contribution of this type of genetic variation to cancer incidence may never be possible because there is a continuum of effects from the imperceptible to the highly significant. Nevertheless it is likely in the long-term that identification of gene - environment interactions in a minority of highly predisposed individuals will have a very significant impact on approaches to prevention and screening. The limiting factor at present is our ability to guess the relevant candidate genes and the carcinogens with which they interact.

In principle, predisposition could occur in one of three ways. The inherited mutation might provide one of the steps in carcinogenesis ready-made in the germ line, make one of these steps more likely to happen, or affect the consequences of one of the steps so as to make the subsequent steps more likely. Most of the inherited cancer syndromes seem to belong to the first category. An example of the second category is xeroderma pigmentosum, in which an inherited defect in DNA repair makes somatic mutation more likely. The variations in carcinogen metabolism described in the last section potentially fall in this group. The third group might include altered endocrine or immune effects on the nascent cancer, but as yet there are no clear examples.

The paradigm for the mechanism of predisposition in the inherited cancer syndromes is retinoblastoma. Since the inherited retinoblastoma mutation is present in every cell of the body, and not every cell becomes a tumour, it is clear that at least one further step is required. The second step might in principle entail either the loss of the second copy of the same gene, or mutation of another gene altogether. In fact, development of the tumour requires loss of both copies of the retinoblastoma gene (Rb). In patients in whom retinoblastoma is non-hereditary both mutations must coincide by chance in the same somatic cell. When the condition is hereditary, the inheritance of one mutation means that it is already present in every cell, which greatly increases the probability that at least one cell will also acquire the second mutation and a cancer will result.

The requirement for loss of the activity of both alleles of the Rb gene implies that the normal activity of the gene is to restrain or suppress tumourigenesis. This concurs with the results from experiments in which hybrids between tumour cells and normal cells are found to be non-tumourigenic, again suggesting that, at the level of the cell, malignancy is the result of recessive mutations in a class of genes that have therefore been called ‘tumour suppressor genes’. The apparent paradox that familial retinoblastoma has a dominant pattern of inheritance, but the mutation is recessive at the level of the cell, is explained because in cancer (unlike other diseases) only one cell need acquire the genetic lesion for the phenotype to be expressed. Thus, although the inherited mutation is recessive in terms of its effect in the cell, only one retinal cell need acquire the second, somatic mutation for a tumour to develop. As a result, almost everyone who inherits the mutation develops the tumour. The phenotype of tumour formation therefore appears to be dominant at the level of the family pedigree.

This would imply that the germline mutation results in loss of activity of the gene and that the second copy of the gene is inactivated in tumours. Definitive proof is not yet available because many of the genes have not been isolated, but in most of the inherited cancer syndromes it seems likely. Some caution is needed because only those syndromes can be discussed for which the chromosomal location of the gene is known. Because the syndromes associated with losses of gene activity (and so, in some cases, obvious chromosomal deletions) are the easiest to map, a spurious impression may arise that all inherited cancers will be of this type.

There is little doubt that in some types of familial cancer, especially the inherited cancer syndromes, screening and appropriate treatment of those at risk will reduce deaths and morbidity. Because the screened population is at very high risk, screening is likely to be highly cost-effective where curative treatment is available. Yet recognition and screening of such families is undoubtedly incomplete and often badly done. Most families are well aware of their history, and worried by it. Even when effective screening or treatment is not available, they may benefit from information and advice. They often overestimate their risks, and can be reassured. The fear of many doctors that they will provoke anxiety by discussing the issues is usually unfounded.

This question reduces to: when is the family history so striking that it is unlikely to be due to chance? Clearly there can be no single answer. The inherited cancer syndromes such as familial polyposis will not usually pose a problem. Two sisters with breast cancer at the age of 40 are more difficult to assess. Such a history may be coincidence, but the figures for excess risk to sisters of cases of this age suggest that there is about an 85 per cent chance that such a pair is ‘significant’. Two sisters with breast cancer at the age of 70, by contrast, may also be ‘significant’, but it is more likely not to be. The factors that suggest genetic predisposition are listed in Table 5 668. In practice, one must take a thorough family history, evaluate these factors, and make an estimate for each new family. The decision will often depend critically on one or two diagnoses in family members now dead. If this is so, considerable effort may be needed to verify the information.

Such clinics have been established in several centres. The aims are first, to obtain as detailed a family history as possible, with confirmation of all important diagnoses; second, to estimate individual risks, discuss these with family members and their doctors, and advise on the options for management; and third, maintain an overview of all the branches of the family and ensure that information from one branch that is relevant to the management of the other is passed on, and that continuity of follow-up is maintained. This is usually beyond the resources of a busy medical or surgical clinic. The purpose of the familial cancer clinic is not to take over clinical management: screening and treatment should remain the responsibility of local specialists, who will see the family regularly.

Management of families with inherited predisposition to cancer requires close liaison with specialist clinicians in disciplines appropriate to each syndrome, such as colonoscopy, ultrasound, mammography, and ophthalmology, as well as with surgeons responsible for prophylactic surgery. It is important to agree on management with the doctors responsible, before advising the patient. Finally, close links with a molecular genetics laboratory are necessary as DNA-based diagnosis becomes possible for more familial cancers, and to foster research.

The first essential is to decide on the risk to the family member(s) concerned. The estimation may involve several steps; the non-specialist may be wise to seek expert help. First, what is the risk that the individual has inherited the predisposing gene, based on a given genetic model (usually autosomal dominant) and his relationship to known affected individuals? Second, might DNA testing give a more precise answer? This is becoming an increasingly important possibility. Up-to-date advice about testing and its limitations should be sought from a clinical geneticist. Third, how is this risk modified by the present age of the individual (the older without signs of the disease, the lower the risk), by any suggestive marker phenotype (e.g. colonic polyps), by the present age (or age at death or prophylactic surgery) of any unaffected relative through whom the gene must have been inherited, or by uncertainty that this family, or branch of the family, does in fact carry a genetic predisposition (e.g. uncertainty whether the supposed gene carrier at the head of the branch might not be a chance case). Consideration must be given to whether there is a risk of more than one type of cancer. If so, which, and what are the risks if different? At what age does risk commence, and how is it distributed through life? How is the risk affected by other known risk factors in this individual (e.g. age at menarche, parity, benign breast disease, etc.).

Having decided on risk, there are generally three possibilities: do nothing, screen for cancer when it is possible, and prophylactic surgery. Sensible advice about lifestyle may well be appropriate. Specific advice on prevention is likely to be a main option in only a few situations at present (e.g. avoidance of sun exposure in familial melanoma).

As a general rule, the patient or family should be helped to make their own decision, rather than have it imposed. The uncertainties of screening and the benefits and drawbacks of surgery must be explained in detail, but at a level appropriate to the individual. The implications for children, even though this may be some years away, is likely to be a major source of concern. It is often helpful to provide parents with a letter detailing the family tree, the discussions that have taken place, and the recommendations for the children, so that this can be passed on for children to use at the appropriate time to seek advice.

Other branches of the family may also be at risk. Whether they will be approached depends primarily on whether they can expect to benefit. If they are not in contact with the branch who have sought advice, or are thought to be unaware of the problem, great caution is needed. Public records in the United Kingdom allow identification of the family doctor, who would be the recommended first contact in these circumstances.

Bussey HJR. Familial polyposis coli. Family studies, histopathology, differential diagnosis and results of treatment. Baltimore: Johns Hopkins University Press, 1975.

Cavenee W, Ponder BAJ, Solomon E, eds. Cancer Surveys Vol 9. Genetics and Cancer. Oxford: Oxford University Press, 1991.

Easton D, Peto J. The contribution of inherited predisposition to cancer incidence. In: Cavenee W, Ponder BAJ, Solomon E, eds, Cancer Surveys Vol. 9. Genetics and Cancer. Oxford: Oxford University Press, 1991.

Hall JM, et al. Linkage of early-onset familial breast cancer to chromosome 17q21. Science 1990; 250:1684.

Knudson AG. Hereditary cancers: clues to mechanisms of carcinogenesis. Br J Cancer 1989; 59:661.

Lynch HT, et al. Hereditary non polyposis colonic cancer (Lynch syndrome I and II). I. Clinical description of resource. Cancer 1988; 56:934.

Malkin D, et al. Germ line p53 mutations in a familial syndrome of breast cancer, sarcomas and other neoplasms. Science 1990; 250:1233.

Murday V, Slack J. Inherited disorders associated with colorectal cancer. Cancer Surv 1989; 8:137.

Ott J. A short guide to linkage analysis. In: Davies KE, ed. Human Genetic Diseases: A Practical Approach. Oxford: IRL Press, 1986.

Ponder BAJ. Prospects for genetic diagnosis of inherited predisposition to cancer. Trends Biotechnol 1990; 8:98.

Ponder BAJ, et al. Risk estimation and screening in families of patients with medullary thyroid carcinoma. Lancet 1988; i:397.

Reik W. Genomic imprinting and genetic disorders in man. Trends Genet 1976; 5:331.

Swift M, Reitnauer PJ, Morrell D, Chase CL. Breast and other cancers in families with ataxia-telangiectasia. N Engl J Med 1988; 316:1289.

Wolf CR. Metabolic factors in cancer susceptibility. In: Cavenee W, Ponder BAJ, Solomon E, eds. Cancer Surveys Vol 9. Genetics and Cancer. Oxford: Oxford University Press, 1991.