The diseases under investigation are assumed to have multiple genetic and nongenetic causes. For simplicity, only the causes involving a single discrete, measurable risk factor and genetic susceptibility are discussed. The risk factor itself may have multiple causes, some of which may involve genetic factors other than the susceptibility genes under study. For illustration, consider the relationship between breast cancer, age at first full-term pregnancy, and genetic susceptibility. Late age at first pregnancy (>30 years) is known to be associated with increased breast cancer risk [Kelsey, 1979
] and may have either nongenetic or genetic origins (e.g., genetic syndromes involving infertility). Several investigators have demonstrated that susceptibility to breast cancer is increased in some families by an autosomal dominant gene [Newman et al., 1988
; Williams and Anderson, 1984
; Go et al., 1983
], but the genetic form of the disease appears to affect only a minority of patients (approximately 4%) [Newman et al., 1988
]. Some women, even in high-risk families, appear to inherit susceptibility from their mothers and transmit it to their daughters without ever becoming affected themselves. This “reduced penetrance” suggests that expression of the dominant susceptibility allele is influenced by environmental factors. What, specifically is the relationship between late first pregnancy and genetic susceptibility? Does the susceptibility gene cause endocrine changes similar to those resulting from delayed pregnancy? Alternatively, does the susceptibility gene operate through a mechanism separate from the effects of delayed pregnancy? If so, does late pregnancy have an effect on gene expression, such as exacerbating it or being required for it?
The five models to be discussed are illustrated in . These illustrations are not path diagrams, but are similar to diagrams used for biochemical pathways. An arrow from one factor (gene, risk factor, or disease) to another indicates that the first factor has a causal influence on the second. An arrow from a factor to an arrow indicates that the factor influences the relationship between the two other factors. When two arrows merge (as in ), it indicates that two factors must both be present to influence disease risk. The five included models are not intended to be exhaustive. Other relationships between risk factors and genetic susceptibility are possible, and may be testable by using the approach described.
Fig. 1 Five hypothetical relationships between genetic susceptibility to disease and risk factors for disease identified in epidemiologic studies. The genetic susceptibility may be either polygenic or due to a dominant, recessive, or X-linked major locus. The (more ...)
In Model A, the genetic susceptibility does not cause disease directly, but acts by increasing the level of expression of the risk factor. In this case, the genetic basis of disease is equivalent to the genetic basis of the risk factor, but the risk factor may have other, nongenetic causes. An example of this model is the relationship between the recessive gene for phenylketonuria (PKU), blood levels of phenylalanine, and mental retardation. Individuals who are homozygous for the PKU gene lack the enzyme necessary to convert phenylalanine to tyrosine, resulting in a buildup of blood levels of phenylalanine [Tourian and Sidbury, 1983
]. These high blood levels, if uncorrected, cause mental retardation. The indirect nature of the effect of the homozygous PKU genotype is illustrated by the prevention of mental retardation if blood phenylalanine is maintained at a low level through dietary intervention. Further, intrauterine exposure to high blood levels of phenylalanine has been shown to cause mental retardation in individuals who lack the high-risk genotype—heterozygous offspring of homozygous PKU mothers [Mabry et al., 1966
In Model B, the risk factor has a direct effect on disease susceptibility, and the genetic susceptibility exacerbates this effect. The genetic susceptibility has no effect in the absence of the risk factor, but the risk factor can act by itself to cause disease. An example of this mechanism is the relationship between xeroderma pigmentosum, ultraviolet radiation, and skin cancer. Individuals with xeroderma pigmentosum have a genetic defect in an enzyme required for repair of DNA damage induced by ultraviolet radiation [Cleaver and Bootsma, 1975
] and are therefore unusually susceptible to sun-induced skin cancer.
Model C is the converse of the second. Here the genetic susceptibility has a direct effect, and the risk factor exacerbates this effect. The risk factor has no effect in the absence of the genetic susceptibility, but the genetic susceptibility can raise risk by itself. Porphyria variegata [Kappas et al., 1983
] is an autosomal dominant genetic disease that fits this model. Affected individuals have skin problems of varying severity, including unusual sensitivity and tendency to blister easily. Upon exposure to barbiturates, however, they experience acute attacks that may lead to paralysis and/or death. Neither the skin problems nor the effects of barbiturates occur in individuals without the gene.
In Model D, neither the genetic susceptibility nor the risk factor can influence disease risk by itself, but risk is increased when both are present. An example of this model is the relationship between the Mediterranean form of glucose-6-phosphate dehydrogenase (G6PD) deficiency, fava bean consumption, and hemolytic anemia [Beutler, 1983
]. G6PD-deficient individuals who consume fava beans develop severe hemolytic anemia, but the disease does not develop either in individuals without G6PD deficiency who eat fava beans or in G6PD-deficient individuals who avoid eating fava beans.
In Model E, either the genetic susceptibility or the risk factor can influence disease risk by itself, and the combined effect of the two may be different from the effect of each acting alone. An example of this model is the relationship between alpha-1-antitrypsin deficiency, smoking, and emphysema [Gadek and Crystal, 1983
]. The disease occurs with elevated frequency in both smokers and individuals with alpha-1-antitrypsin deficiency, but smokers who also have the enzyme deficiency have even more dramatically elevated risk.