Most common human psychiatric and medical disorders such as schizophrenia, diabetes, and heart disease do not show a Mendelian pattern of inheritance. Neither can they be considered as continuous traits in that people are described as being affected or unaffected. However, these conditions could be regarded as quasi-continuous in that those who are affected can be graded along a continuum of severity. It is possible to extend this to assume that there is an underlying liability to develop the disorder which is continuously distributed in the population. Those who pass a certain threshold manifest the condition. If the underlying liability to develop the disorder is inherited in a polygenic or multifactorial fashion, then we can assume that the distribution will be approximately normally distributed (Fig. 2.4.1.2). The genetic liability of relatives of affected individuals will be increased and their liability distribution will be shifted to the right (Fig. 2.4.1.2). Thus, the proportion of relatives above the disease threshold will be greater compared with the general population. If we know the proportion of affected relatives of probands and the proportion of those affected in the general population, it is possible to calculate the correlation in liability between pairs of relatives using this type of model.

An alternative to a polygenic model of complex disease is a single major locus model. Single-gene disorders do not always show typical Mendelian patterns of inheritance. For example familial transmission can be modified by variable expression and penetrance. Some conditions can show great variability in terms of clinical expression. For example neurofibromatosis, an autosomal dominant disorder can express itself as the full blown disorder or merely as a few café-au-lait spots. Penetrance is defined as the probability of manifesting the disorder given a particular genotype. For Mendelian disorders this is always 1 or 0, but irregular patterns of inheritance may occur because of incomplete penetrance where the probability of manifesting the disorder is greater than 0 but less than 1.

Finally, there are now molecular explanations for other types of unusual patterns of inheritance for single genes. Anticipation where disorders show a progressively earlier onset and greater severity with subsequent generations is now known to be explained by heritable unstable nucleotide repeat sequences (see later). Huntington’s chorea and fragile X syndrome are examples of disorders caused by heritable unstable repeats. For some conditions such as the Angelman syndrome and the Prader-Willi syndrome, manifestation of the disorder depends on the parental origin of the gene. This is known as imprinting.

Alternative explanations of how complex conditions such as psychiatric disorders are inherited include mixed and oligogenic patterns of transmission. A mixed model includes a major gene and a polygenic/multifactorial contribution. However, for many of these disorders, genes of major effect may not exist. It may be that these irregular phenotypes are best explained by oligogenic models where the co-action or interaction of a small number of genes contributes to the disorder. These issues remain to be resolved by molecular genetic studies (see later).

Family studies allow us to determine whether a disorder aggregates in families by examining the rate of disorder in the relatives of affected individuals (probands) and comparing this with the rate of disorder in the general population or in a control group. Alternatively we can compare the frequency of disorder in the relatives of probands with the frequency among relatives of a control group of normal individuals or those with another disorder.

There are two types of family studies. The family history method is more economical in that the psychiatric history is taken from the proband. However, given that most individuals are unlikely to know as much about family members as about themselves, this method results in an underestimate of diagnoses in relatives. A more thorough but more time-consuming approach is the family study method where all available relatives are directly interviewed.

An important issue is how a family study sample is ascertained. Ideally, probands should be ascertained independently from each other. This is unlikely to pose a problem for rare disorders. However, for more common conditions where a series of cases is collected, for example, by consecutive referrals of the disorder to a particular hospital, it is possible that families included in a family study contain more than one proband. This is known as multiple incomplete ascertainment. Complete ascertainment, where all affected individuals in a given population are included, is rarely possible and in most instances probands are identified after some selection process (e.g. referrals to a particular hospital). Thus, factors influencing selection, such as comorbidity and help seeking may also influence observed patterns of familial aggregation.

For genetic studies, we are interested in the proportion of individuals who have ever had the disorder (lifetime prevalence) rather than the proportion who show the disorder at one point in time (point prevalence). However, a difficulty encountered when carrying out family studies is that the observed rates of disorder will also depend on the age of the individual, the risk period for the disorder, and whether or not the individual has lived through that risk period. Thus, some members may not yet have reached the age of risk for the disorder, some are currently unaffected but will become affected at some later point, and others may have died whilst still unaffected. The most appropriate method is to correct for age and express the rate of disorder in relatives as the morbid risk (MR) or lifetime expectancy.

There are many methods of age correction, of which the Slater-Stromgen adaption of Weinberg’s shorter method is the most straightforward. The MR of the disorder can be estimated as the number of affecteds (A) divided by the bezugsziffer (BZ) where the BZ is calculated as:

Σ_i[a1]w_i + A

and where w_i is the weight given to the ith unaffected individual based on their current age. The most accurate approach is to use an empirical age of onset distribution from a large separate sample, for example a national registry of psychiatric disorders, to obtain the cumulative frequency of disorder over a range of age bands, from which weights can be derived.

Another approach is to carry out life table analysis. The distribution of survival times (or times to becoming ill) is divided into a number of intervals. For each of these one can calculate the number and proportion of subjects who entered the interval unaffected and the number and proportion of cases that became affected during that interval as well as the number of cases that were lost to follow-up (because they had died or had otherwise ‘disappeared from view’). Based on these, the numbers and proportions ‘failing’ or becoming ill over a certain time interval (usually taken as the entire period of risk) can be calculated. A further alternative is to use a Kaplan Meier product limit estimator. This allows one to estimate the survival function directly from continuous survival or failure times instead of classifying observed survival times into a life table. Effectively this means creating a life table in which each time interval contains exactly one case. It therefore has an advantage over a life table method in that the results do not depend on grouping of the data.

One potential source of bias in twin studies stems from ascertainment procedures. For example, affected twins referred to a specific study or volunteer samples are likely to include more twin pairs who are monozygotic and who are concordant. Ascertainment of twin pairs through hospital registers overcomes this problem to some extent, but for some disorders may be biased by the process of referral. Population-based samples overcome these biases, although when examining disorders rather than traits extremely large sample sizes are required to obtain an adequate number of affected individuals.

A further potential source of error is in the assignment of zygosity. Ideally zygosity should be determined by DNA typing. However, it may be more practical to use a twin similarity questionnaire which includes questions such as whether the twins share the same hair/eye colour, and whether they look alike as two peas in a pod. This method of assigning zygosity is simple and inexpensive with a reported accuracy of over 90 per cent.

It is sometimes argued that a major drawback to the twin study method is that monozygotic twins may experience a more similar environment and may be treated more similarly than dizygotic twins. However, where there is evidence that monozygotic twins share greater environmental similarity than dizygotic twins it is difficult to infer whether this contributes to their similarity for the disorder or whether this is the consequence of greater genetic similarity. There have been several approaches adopted to further explore this issue.

In some studies questionnaire measures of environmental sharing (e.g. being dressed alike as children, sharing friends) have been used. These suggest that environmental sharing is indeed greater for monozygotic twins than for dizygotic twins. However, it appears that for many traits and disorders such as cognitive ability, personality, depressive symptoms, and depressive disorder this degree of similarity for childhood environment does not account for monozygotic twin similarity for the trait. One way of disentangling cause and effect is to use direct observational studies. Although this method has not been much used, one study of young twins suggested that the greater similarity of parental responses to monozygotic twins compared to dizygotic twins appeared to be elicited by the twins themselves.

An alternative method of examining the effects of environmental sharing is to study twins who are mistaken about their zygosity. However, most studies which have used this method suggest that perceived zygosity is a less important influence on twin similarity than true zygosity.

Finally, the most powerful means of examining the effects of environmental sharing is to look at monozygotic twins who have been reared apart. However, such twin pairs are rare and have mostly been ascertained in a biased fashion. Nevertheless, studies of reared-apart twins have informed us that there is a substantial genetic contribution to cognition, personality, and psychosis.

The final potential criticism of the twin method is whether twins can be regarded as representative of the general population given some important differences. Twin births are relatively common (1 in 80 births), although the number of dizygotic twins varies in different countries and is influenced by factors such as maternal age and multiparity, a family history of twins and increasingly, the use of fertility drugs. Twins are more likely to experience greater intrauterine and perinatal adversity and the experience of being brought up as a twin is unusual in itself. There is also some evidence that depression is more common in mothers of young twins than among mothers of singletons. However, these differences are only important if they result in different rates of disorder or symptoms in twins compared to singletons. So far there is little evidence to suggest that the rate of psychiatric disorder in twins is any higher than amongst singletons.