Increased Rate of Twins among Affected Sib Pairs

To the Editor:

Recently, Greenberg et al. (2001) and Betancur et al. (2002) reported an excess of twin pairs among affected sib pairs with autism (MIM 209850). Greenberg et al. (2001) reported an excess of both MZ and DZ pairs, whereas Betancur et al. (2002) found an excess of MZ pairs only. Both studies tested the rates of twin pairs among a sample of affected sib pairs against the population rates. The hypothesis put forward was that being a twin is in itself a risk factor for autism. The purpose of this letter is to show that an excess of twin pairs among affected siblings—in particular, an excess of MZ pairs—is what would be expected if genetic factors are implicated in the etiology of a disorder and does not in itself suggest that being a twin confers a risk. Hence, the reported results could be a logical consequence of the affected sibling ascertainment scheme.

The proportion of twin pairs among a random sample of affected siblings from the population depends on the population incidence of twinning and on the concordance rate for the disorder. Let p be the incidence of the disorder in the population; f_MZ and f_DZ the population rates of MZ twins and DZ twins, respectively; and r_S, r_DZ, and r_MZ be the (casewise) concordance rates (i.e., the probability that one sibling is affected, given that the other sibling is affected) for nontwin siblings, DZ, and MZ twins, respectively. For each of the three kinds of sib pairs, the probability of 0, 1, and 2 affected individuals is, for r=r_S, r_DZ, r_MZ,

It follows that the proportion of MZ pairs among all pairs of affecteds is

Note that this proportion is independent of the population incidence. For small DZ and MZ population rates, f^*_MZ≈f_MZr_MZ/r_S; that is, we would expect an increase in the rate of MZ twins that is proportional to the increase in the concordance rate relative to nontwin siblings. From epidemiological studies, the estimates for the concordance rates for autism in MZ pairs, DZ pairs, and nontwin siblings are approximately 0.4–0.7, 0.0–0.03, and 0.03, respectively (see Lauritsen and Ewald [2001] and Folstein and Rosen-Sheidley [2001] for reviews), consistent with a very high heritability on a liability scale and the existence of nonadditive genetic variation for liability (see, e.g., Smith 1970). These estimates suggest that the proportion of MZ twin pairs in a random sample of affected sib pairs is approximately 13–23 times larger than the population MZ twinning rate. The observed increases in the MZ rate in the Greenberg et al. (2001) and Betancur et al. (2002) reports are 13 and 16, respectively; they are in accordance with the published concordance rates.

Greenberg et al. (2001) also report a significant increase (a nearly fivefold increase) in the proportion of DZ twins among the affected sib pairs. Estimates of DZ concordance rates have been similar to or lower than the rates among nontwin siblings but have been based on small numbers of observations (Folstein and Rosen-Sheidley 2001; Lauritsen and Ewald 2001). An increase in the rate of DZ twins relative to nontwin siblings could be due to common environmental factors or due to the “stoppage” phenomenon, in which parents with one affected child choose not to have more children. Lastly, Greenberg et al. (2001) compare their observed increased rates of autism in affected twin pairs with the rates for insulin-dependent diabetes mellitus (IDDM). They found a deficit of DZ twin pairs but an excess of MZ twin pairs. These results are also consistent with the genetic epidemiology of IDDM, with reported concordance rates of 0.06, 0.11, and 0.30–0.50 for nontwin siblings, DZ twins, and MZ twins, respectively (see, e.g., Kyvik et al. 1995; Field 2002).

In this letter, I have suggested another explanation for the observed excess of twin pairs among affected sibling pairs; that it is simply the effect of ascertaining pairs of affected siblings. Is multiple birth an important risk factor for autism? The data presented by Greenberg et al. and Betancur et al. do not allow the testing of this hypothesis. A population-based study, in which the incidence of autism among MZ twins, DZ twins, and nontwin siblings is estimated should clarify this important issue.