36, 310–333 (1998) BR970972 BRAIN AND COGNITION ARTICLE NO. A Twin and Family Study of the Association Between Immune System Dysfunction and Dyslexia Using Blood Serum Immunoassay and Survey Data Jeffrey W. Gilger,* Bruce F. Pennington,† Ronald J. Harbeck,§ John C. DeFries,** Brian Kotzin,‡ Phyllis Green,† and Shelley Smith*** *University of Kansas; †University of Denver; §National Jewish Center for Immunology and Respiratory Medicine, Denver; **University of Colorado, Boulder; ‡University of Colorado, Denver; ***Boys Town National Research Hospital We conducted a study of the association between developmental reading disability (DRD) and immune disorders (ID) using both survey and immunoassay data in two separate samples of families. One sample was made up of twins and their parents and was ascertained through a population-based sampling scheme. The other sample was a set of extended pedigrees selected for apparent autosomal dominant transmission of DRD. We failed to find an association between DRD and ID in either sample, regardless of the method used to assess immune system function. Even though our twin sample provided evidence that both DRD and immune conditions were significantly heritable, there was no evidence for a genetic correlation between ID and DRD nor was there any clear indication that a special subgroup of individuals may be comorbid for these conditions because of genetic reasons. How these negative findings can be reconciled with the developmental hypothesis of Geschwind, Behan, Galaburda, and colleagues, and how they may relate to the gene locus influencing DRD that has been recently located in the HLA region of the short arm of chromosome 6 is discussed.  1998 Academic Press Geschwind and colleagues (Geschwind & Behan, 1982, 1984; Geschwind & Galaburda, 1985a, 1985b) suggested that phenotypic associations between learning disabilities (e.g., developmental reading disorder or DRD), immune disorders (ID), and nonrighthandedness may occur because all three traits have a common etiology. Even though the mechanisms postulated for these associations are complex, they essentially involve the prenatal effects of testosterone on the developing brain and thymus gland. Because of their Funded in part by a grant from the Rita Rudel Foundation (J. Gilger), and a Learning Disabilities Center Grant P50 HB27802. Bruce Pennington was also supported by a Research Scientist Award, K02 MH00419, and a MERIT award, R37 MH38820. Send reprint requests to J. Gilger, University of Kansas, 3031 Dole Human Development Center, Lawrence, KS 66045, or e-mail gilger@falcon.cc.ukans.edu. 310 0278-2626/98 $25.00 Copyright  1998 by Academic Press All rights of reproduction in any form reserved. TWIN AND FAMILY 311 common etiology, associations among ID, DRD, and handedness should cluster in families, although these associations may not be seen consistently in individuals. Thus, the three phenotypes can be viewed as a consequence of the pleiotropic effects of an underlying common etiologic factor(s) that is (are) familial, but not necessarily genetic. As noted in a recent review (Bryden, McMannus, & Bulman-Flemming, 1994) and subsequent commentaries (Brain and Cognition, 26, 1994; 27, 1995) there have been literally hundreds of studies aimed at some facet of the hypothesized links among these three conditions. Bryden et al. (1994) pointed out that with few exceptions, the results across studies are inconsistent, and that even the theoretical basis for the associations may be in error. Many of the tests for the hypothesized associations have methodological problems, such as biased ascertainment of subjects, inadequate measures of phenotypes, and small sample size. Further, most studies have used data on individuals, rather than families, to examine pairwise associations (e.g., rates of ID in children with dyslexia) and have not assessed the hypothesized familial aggregation and cosegregation (however, see Gilger, Pennington, Green, Smith, & Smith, 1992, and Urion, 1988). The recent finding of a possible quantitative trait locus (QTL) influencing DRD located in the human histocompatibility antigen (HLA) region on the short arm of chromosome 6 (Cardon, Smith, Fulker, Kimberling, Pennington, & DeFries, 1994; Grigorenko, Wood, Meyer, Hart, Speed, Shuster, & Pauls, 1996) raises the possibility that the DRD–ID association found in some studies might be genetically mediated. The HLA region is highly polymorphic and contains many genes that influence immune function. It is possible that this QTL either pleiotropically affects immune function as well as DRD or is linked to a separate immune system gene. In either case, cosegregation of DRD and immune disorders (that are a reflection of HLA genes or other genes on chromosome 6; e.g., Bias, Reveille, Beaty, Meyers, & Arnett, 1986; Fielder, Walport, Batchelor, Rynes, Black, Dodi, & Hughes, 1983) in at least some DRD families could result (Gilger & Pennington, 1995; Gilger, Pennington, & Defries, 1992). The purpose of this study is threefold. First, our goal is to test for the hypothesized (i.e., according to the Geschwind–Behan–Galaburda, or GBG model) phenotypic association between DRD and ID using improved methods of diagnosis. This includes actual standardized tests of reading ability in all subjects, detailed survey information on immune system function, and supplementary blood serum immunoassay data that may help identify presymptomatic individuals. Second, we use twin and family data that will allow more than simple pairwise tests of phenotypic association: we also test for a genetic etiology of the hypothesized ID–DRD link. Cosegregation or cotransmission in relatives is implied by the GBG theory because ID and DRD are postulated to share a common familial etiology. Thus, for example, the presence of DRD in a proband should place his or her relatives at an in- 312 GILGER ET AL. creased risk for ID.1 This may occur only for a subset of families, suggesting a DRD-ID subtype in the population (Gilger & Pennington, 1995). Third, we analyze two large and distinct samples, providing a test of generalizability. Moreover, part of our sample includes subjects in which a QTL influencing DRD in the HLA region was found (Cardon et al., 1994; Grigorenko et al., 1996). Finally, at least one of our samples, the twins, are ascertained in a manner consistent with a population based scheme. Consequently, many of the potential biases and artifacts that may arise from the commonly used and highly select clinical samples of past research can be avoided. METHODS There are two independent samples in the present study: Twin pairs from the Colorado study of reading disabilities and family members from the linkage kindreds. For a subset of the Colorado twins, parent and sibling data were also collected. The Colorado Twin Study Subjects. As of July 1996, the ongoing Colorado Twin Study of Reading Disabilities had collected data on approximately 222 monozygotic (MZ), and 281 same-sex dizygotic (DZ) twin pairs, where at least one member of each pair is reading disabled (DeFries, Filipek, Fulker, Olson, Pennington, Smith, & Wise, 1997). A control group of 148 MZ and 160 samesex DZ twin pairs selected through nonreading disabled probands has also been obtained. Control probands are matched to affected proband twins on the basis of age, sex, and school district. All twins resided in Colorado at the time their data were collected. As part of the study, all twins were administered large test batteries that assess cognitive, reading, and language abilities (see DeFries et al., 1997, and Pennington, Gilger, Olson, & DeFries, 1992, for a complete description). Information was also obtained to ensure that subjects had no diagnosed neurological, emotional, or uncorrected visual or auditory problems that might be responsible for their learning problems. All DRD twins were required to have a positive school history of reading problems (e.g., teacher or counselor reports, poor grades, test scores) and either verbal or performance scale IQs in the normal range. The parents of potential twin pairs were contacted by letter and phone, and informed consent was obtained for all subjects. Materials. A variety of questionnaires were administered to the parents of twins. The surveys provided information used in the diagnosis of zygosity (Nichols & Bilbro, 1966), as well as reading history, medical and socioemotional status, and other topics. Among the cognitive tests administered to twins were the Peabody Individual Achievement Test (PIAT; Dunn & Markwardt, 1970), the Wechsler Intelligence Scale for Children—Revised (WISCR; Wechsler, 1974) or the Wechsler Adult Intelligence Scale-Revised (WAIS; Wechsler, 1981), and a variety of measurements of specific cognitive abilities such as perceptual speed, memory, and spatial skills. All test measures are commonly used and possess acceptable validity and reliability. The zygosity questionnaire is also adequately valid and reliable, having an overall accuracy rate that exceeds 90% (Nichols & Bilbro, 1966). Several years after the inception of the Colorado twin study, a decision was made to re1 For this paper, we have chosen to focus only on the association between ID and DRD, and not to consider handedness. We have already addressed handedness for these samples in Gilger, Pennington, Green et al. (1992), and our new data do not include any additional information pertinent to the handedness variables. TWIN AND FAMILY 313 contact as many of the twin families as possible and obtain more detailed information on subjects’ medical historys for various immune disorders (e.g., allergies, autoimmune diseases, asthma, etc.). The overall response rates for the additional medical history surveys approximated 75% of the recontacted sample. Although data for some pairs are incomplete, at the time of this writing there was information on immune disorders for approximately 846 individuals from DRD proband pairs (i.e., 423 twin pairs) and 496 individuals from control proband pairs (i.e., 248 twin pairs). Parents of twins also answered immune surveys about themselves. Diagnosis of DRD. After selection and testing, twins were diagnosed as reading disabled by a discriminant function that uses weighted scores from the PIAT Reading Recognition, Reading Comprehension, and Spelling subtests. The discriminant weights used were estimated from an earlier analysis of data from an independent sample of 140 nontwins referred for reading problems and 140 control children. In order for a member of a twin pair to be diagnosed as reading disabled, the individual must be classified as affected by the discriminant function (i.e., a discriminant score less than or equal to 0). The discriminant function is essentially a normative-age-based criterion, and it has been shown to be a valid diagnostic method and a sensitive reading phenotype (e.g., Pennington et al., 1992; see also Fletcher, 1992, for a discussion of age-based versus other DRD diagnostic schemes). For example, it demonstrates false positive and false negative rates similar to IQ discrepant formulas, differentially correlates with phonological and orthographic skills (with higher correlations with phonological processing), and appears to tap a basic genetic etiology expressed as a DRD (Pennington et al., 1992; Cardon, DeFries, Fulker, Kimberling, Pennington, & Smith, 1994; Pennington, Gilger, Pauls, Smith, Smith, & DeFries, 1991). Furthermore, when used as a standard of comparison for identifying compensated adults (i.e., those with DRD as children, but not as adults according to discriminant score cutoffs), the discriminant diagnosis differentially predicts familial risk for DRD in offspring of compensated and noncompensated parents, again suggesting that it is sensitive to an underlying familial etiology of DRD (Gilger, Hanebuth, Smith, & Pennington, 1996). Descriptive statistics. Descriptive statistics for the twin sample are presented in Table 1. Sample sizes may deviate from those presented elsewhere in this paper because data are incomplete for some pairs or individuals. Within the DRD proband sample, the mean IQ of the non-DRD subjects significantly exceeds that of the DRD subjects, even though both groups have WISCR IQs in the normal range. Table 1 also shows that DRD subjects of these pairs performed significantly lower on the weighted PIAT subtests that make up the discriminant score. The male-to-female ratio for the diagnosis of DRD in the affected twin sample is 1 : 1, and this deviates from ratios typically seen in school and clinic-referred samples (i.e., 3 : 1). In part, we attribute this 1 : 1 ratio to the sex bias often present in volunteer twin studies, where more females than males typically participate (Lykken, Tellegen, & DeRubeis, 1978). We also suggest that this sex ratio is a more accurate representation of the actual sex ratio for DRD: prior work has shown similar sex ratios when samples are population-based or otherwise corrected for ascertainment biases (e.g., Shaywitz, Shaywitz, Fletcher, & Escobar, 1990; DeFries, 1991). Subset of Twins from the Twin Family Study Subjects. Subsequent to psychometric testing, a small subset of the twins was recruited to have blood taken for our immunoassay work. Our purpose was to obtain a twin sample for immunoassay reliability checks for our linkage family data (see the discussion that follows) and for a small sample replication of any significant results yielded by the linkage family data. A trained phlebotomist went to the families’ homes to draw blood. At that time, informed consent was obtained along with further detailed medical history of autoimmune and allergic illnesses. Each family member was paid $10.00 for participation. Ten MZ and 10 DZ twin families where at least one member of each pair was DRD agreed to have blood samples 314 GILGER ET AL. TABLE 1 Descriptive Statistics and F-Tests for DRD and Non-DRD Members of the Colorado Twin Studya Variable DRD non-DRD Twins Pairs Selected Through DRD Probands N individuals 519 Age (years) 12.44 (5.68) WISCR FSIQ 97.14 (10.44) Discriminant score 21.12 (0.84) Male : female DRD ratio 5 1.0: 1 288 13.87 (8.48) 104.26 (10.41) 0.76 (0.61) Twins from DRD and Control Pairs b N individuals 846 Age (years) 11.44 (2.59) WISCR FSIQ 99.68 (11.0) Discriminant score 20.46 (1.20) Male : female DRD ratio 5 1.0: 1 496 11.92 (2.83) 113.00 (10.60) 1.29 (0.86) Subset of Twins c N individuals Age (years) 40 14.8 (2.11) WISC-R FSIQ 101.83 (11.43) Discriminant score 20.197 Male : female DRD ratio 5 1.96 : 1 N individuals 33 Age of parents (years) 44.7 (5.99) 24 14.67 (2.01) 116.00 (12.04) 1.81 22 40.9 (5.89) F 8.04 p ,.01 124.61 ,.001 1111.45 ,.001 9.89 ,.01 649.12 ,.001 809.56 ,.001 ,1.0 ns 22.17 ,.001 43.17 ,.001 5.47 ,.05 a Data presented under DRD and non-DRD columns are means and standard deviations in parentheses. b Data for affected probands and their cotwins are under the DRD column, and data for control pairs of twins are under the non-DRD column. c Data for affected proband pairs of twins are under the DRD column, and data for control proband pairs of twins are under the non-DRD column. See text for explanation of subset selection. taken. Seven MZ and 5 DZ twin families from the non-DRD group also had blood samples drawn. The DRD and non-DRD twins were matched as a group on gender, age, zygosity, and geographic location. In all cases, both twins and at least one parent were drawn. No siblings were drawn. Descriptive statistics. Table 1 also contains data from the subset of twin families. Complete data were collected on 119 people in the subset of twin families. As with the main twin sample, the mean Weschler FSIQ for the non-DRD twins is significantly higher than the DRD twins. TWIN AND FAMILY 315 The discriminant function score also was significantly lower in the DRD group. The parents of DRD subjects were significantly older than the parents of non-DRD control subjects. Although not shown, the ratio of mothers and fathers sampled was roughly the same for both twin groups. The Linkage Kindreds Subjects. About 20 years ago, a collaborative study was begun that included a linkage analysis of families selected through a dyslexic proband (Smith, Kimberling, Pennington, & Lubs, 1983). Probands were ascertained through schools, tutors, and parent groups, and only those pedigrees suggestive of autosomal, major gene transmission (e.g., a three-generation history of familial reading problems) of dyslexia were asked to participate. For this paper, data are available on approximately 272 blood relatives from 19 three-generation kindreds. Although family members reside throughout the United States, roughly 50% are located in Colorado and the Western plains states. Materials. Subjects were tested and interviewed by trained personnel, and tests and questionnaires were given to child and adult subjects (see Smith et al., 1983, and Pennington et al., 1987, for a complete list of tests administered). Survey data on children were provided by their parents, most often the mother. Among the cognitive tests were the Raven’s Progressive Matrices and the PIAT. Other tests of cognitive ability were administered as well. The surveys addressed reading history, medical, socioemotional, and other general information. A phlebotomist drew blood samples from the individuals in these kindreds for the linkage analyses. After DNA extraction for the original linkage analyses, the remaining blood was later used for the immunoassays. Diagnosis of DRD. For this paper, subjects are diagnosed as dyslexic or nondyslexic in the same fashion as in the Colorado twin sample (for details see earlier discussion). Specifically, a subject is diagnosed as reading disabled by a discriminant function that uses weighted scores from the PIAT Reading Recognition, Reading Comprehension, and Spelling subtests. A discriminant score less than or equal to 0 is required to be classified as dyslexic. Descriptive Statistics. Descriptive statistics for the linkage sample are shown in Table 2. Sample sizes may deviate from those presented elsewhere in this paper because test data were not obtained from all relatives (e.g., some adults refused to be tested but provided history data and blood samples). Excluding probands, there were sufficient test data to diagnose approximately 107 blood relatives as reading disabled and approximately 135 blood relatives as nonreading disabled, using the discriminant score method. The mean Raven’s IQ is significantly larger for nonDRD subjects, although the mean IQs of both groups are in the normal to above normal ranges. The male-to-female ratio for DRD in both groups is 1 : 1. Similar to the twin sex ratio presented, we attribute this equal sex ratio to the correction for clinical ascertainment that essentially occurs when the probands are not included in the analysis. In fact, as Table 2 shows, the sex ratio for the probands was 3 : 1, and this is in close accordance with expectations based on clinical samples. The data presented in Table 2 also show that the probands are of normal intelligence and, as expected, show greater deficits in the discriminant score compared to both DRD and non-DRD relatives. PROCEDURES Survey Information Immune data were systematically gathered in the twin and linkage studies via questionnaire. To be counted as having an atopic or autoimmune disorder, subjects needed to note that the condition was diagnosed by a physician. 316 GILGER ET AL. TABLE 2 Descriptive Statistics and F-Tests for DRD and Non-DRD Proband Blood Relatives of the Linkage Data Seta Variable DRD non-DRD Blood Relatives of Affected Kindreds (excl. probands) N 107 138 Age (years) 30.03 33.05 (16.54) (18.29) Raven’s IQ 111.34 117.69 (11.74) (9.31) Discriminant score 21.35 1.14 (1.16) (0.82) Male : female DRD ratio 5 1 : 1 F p 1.79 ,.05 17.69 ,.001 384.73 ,.001 Probands N Age (years) 19 18.86 (8.97) Raven’s IQ 109.80 (11.17) Discriminant score 21.62 (1.74) Male : female DRD ratio 5 3.0: 1 a Data presented under DRD and non-DRD columns are means and standard deviations in parentheses. Subjects were excluded from the analyses if physician diagnosis of their disorder was unclear or questionable. This method is common to most studies that have examined the GBG hypothesis. Even though it may appear that the requirement for a formal diagnosis may tend to underestimate the frequency of immune disorders, the rates reported by subjects are consistent with the rates estimated to exist in the population. In fact, especially in terms of atopias, the rates we obtain seem a bit higher than expected (see the following discussion). An individual was classified as having an autoimmune disorder if he or she was positive for any of the following medically documented conditions: early onset rheumatoid arthritis; multiple sclerosis; thyroid disorders (e.g., Hashimoto’s, Grave’s); ulcerative colitis; systemic lupus erythematosus; early onset diabetes mellitus; myasthenia gravis; uveitis; or dermatomyositis. Allergic or atopic conditions were identified in subjects reporting medically documented hay fever, asthma, or food or skin reactions. The reader should be aware that because many of the subjects in our studies will not have passed the age of risk for some of these disorders, there is a conservative bias in these data for many of these conditions.2 2 These families and twins are part of a continuing project and will be followed for years to come. Future analyses will examine if later onset of immune disorders affects the results we report here. TWIN AND FAMILY 317 The distribution of responses to the autoimmune questions were as follows: in the linkage families, 90% of the individuals reported no problems, 7% reported one problem, 2% reported two problems, and 0.03% reported four problems; in the twin families (including twins and all parents that returned the surveys), 97% reported no problems, 3% reported one problem, 0.03% reported two problems, and three problems were reported by one individual. The allergic conditions were distributed as follows: in the linkage families, 74% reported no problems, 19% reported one problem, 7% reported two problems, 1% reported three problems, and one individual reported four problems; in the twin families, 72% reported no problems, 20% reported one problem, 5% reported two problems, 2% reported three problems, 0.08% reported four problems, and five problems were reported for 0.02% of the individuals. For both autoimmune and atopic conditions, a single composite score was obtained by counting how many conditions each person reported. If an individual reported at least one autoimmune condition, he or she was coded as 1 for the summary variable AUTO but as 0 otherwise. Similarly, he or she was coded as 1 for a summary variable ALLERGY if there was at least one report of an allergic condition, and as 0 otherwise. Preliminary analyses not reported here suggested that these summary variables were an adequate way to represent our data. Specifically, doing distinct analyses on each separate autoimmune or allergy survey item was not reasonable because of the resultant sample size and statistical problems, and our preliminary analyses suggested that there were no particular variables that better distinguished individuals or groups than others. Immunoassay Data The antinuclear antibody (ANA) and rheumatoid factor (RA) assays were performed in a standard fashion in coauthor Harbeck’s laboratory, which is accredited by the College of American Pathologists. These are standard clinical tests used in the evaluation of patients with symptoms of immune-related disorders. An indirect fluorescent antibody technique (Quantafluor, Kallestad Diagnostics, Chaska, MN 55318) using HEp-2 cells was employed to determine the level of antinuclear antibody in a sample (Fritzler, 1992). Serum samples were run at an initial 1 :40 dilution, and positive sera were titered to endpoint (last dilution which yielded a fluorescence intensity of 11). The successively higher titer dilutions were 1/160, 1/320, 1/640, and 1/1280. Samples with a 11 fluorescence intensity at a dilution at or above 1/80 were coded as positive for the main analyses. Rheumatoid factor concentration in subjects’ sera was measured using a nephelometric technique (Cook & Agnello, 1992). In brief, polystyrene particles coated with a human gamma globulin antihuman gamma globulin complex (Behring Diagnostics, Inc., Marburg, Germany) are mixed with the sub- 318 GILGER ET AL. ject’s serum. The intensity of the resulting scattered light as measured in a nephelometer (Behring Diagnostics, Inc.) is dependent upon the RA content. Hence, comparison to standards of known concentration allows one to determine the RA content of the sample. The limit of sensitivity with this method is 20 IU/ml. The RA test is used and interpreted cautiously in the diagnosis of rheumatoid arthritis (Heimer, Levin, & Rudd, 1963; Mikkelsen, Dodge, Duff, & Kato, 1967). In a large study of normal individuals, 3.3% had positive RA tests, and this ‘‘false positivity’’ increases with age and has been shown to be as high as 40% for individuals over 75 years of age. Thus, it is advised that independent laboratories evaluate and determine the normal ranges for the population typically seen, as was done in Harbeck’s laboratory. Similarly, for the ANA assay using a HEp-2 substrate, the rate of positivity in an apparently normal sample has been reported to be as high as 28.7% at a 1/40 dilution, so oversensitivity of this assay is also a concern (Lipscomb, Cope, Stephens, Deng, & Gilliam, 1984; Sontheimer, 1993). Even at a cutoff of 1/160, there is a 5% positivity rate in normal adults in their 30s, which rises to 17% for normal adults in their 60s. As shown in the Results section, we found higher than expected rates of positivity in our samples for both the ANA and RA assays. Analyses Several analyses were conducted: (1) exploratory/preliminary analyses, which are concerned with assortative mating, and relations between the immune variables and age and sex; (2) information on the reliability and validity of immunoassay data relative to the survey data and expectations given prior population samples; (3) tests for associations between DRD and ID; and, (4) tests for a common etiology between DRD and ID. The analytical methods used to address the last group of tests are described next. Tests for a Genetic Etiology of an ID–DRD Association. In Gilger, Pennington, & DeFries (1992), we performed cosegregation analyses on three sets of families (including the present linkage sample) and on a subset of the twins we present in this paper to test for a common etiology of DRD and ID. All results were null regarding common genetic etiology in the samples overall and for a genetically mediated subtype of DRD–ID individuals. Because the twin sample is much larger now, we will repeat our tests for a genetic association in this sample only. Because of the small base rates and the limited amount of immunoassay data for twins, cosegregation analyses are not possible for DRD and autoimmune disorders and immunoassays, and therefore we will conduct analyses only for DRD the ALLERGY summary variable. There are a variety of ways to analyze bi- or multivariate twin data for joint genetic effects (e.g., Hannah, Hopper, & Mathews, 1983; Hopper, Hannah, TWIN AND FAMILY 319 Macaskill, & Mathews, 1990; Neale & Cardon, 1992). However, because of their assumptions, these techniques may not be well suited for use with selected samples like the Colorado DRD twins. Moreover, using simple bivariate cross-concordances can be a quite elegant and straightforward approach to assessing the shared genes between two qualitatively measured traits such as DRD and ID. We will perform two types of analyses: one that tests for a common etiology in the entire twin sample for ID and DRD and another that tests for an etiologic subtype. In the first analysis, all MZ and DZ pairs are compared in terms of the degree of cross-concordance for DRD and ID. Specifically, the method selects twin pairs where at least one member is DRD and examines the rates of ID in the cotwins. A significantly higher MZ than DZ crossconcordance suggests some degree of a common genetic etiology for DRD and ID. Because of the possibility of etiologic heterogeneity for any DRD–ID comorbidity observed, a subtype analysis is also conducted. In a two disorder model, such as DRD and ID, there are four possible ways a person may be affected: he or she may have both disorders (11), just DRD (12), just ID (21), or neither disorder (22). It is possible that these subtypes are etiologically distinct. The best chance of finding a genetically mediated subtype where DRD and ID are coinherited is to select a subset of DZ pairs, where at least one member of the pair is 11.3 A comparison can then be made of the frequency of DZ cotwins also being a 11 phenotype, rather than the other three phenotypes possible (12, 21, 22). A higher DZ concordance than expected by chance would suggest a genetic component to DRD–ID comorbidity in some subpopulation of people. The rates in DZs alone rather than in comparison to MZs are examined because members of DZ pairs, like other family members other than MZ twins (that share 100% of their genes), can be used to assess how segregating genes in common manifest themselves in similar phenotypes. Analogous to the family cosegregation studies in Gilger, Pennington, Green et al. (1992) and Pauls, Hurst, Kruger, Leckman, Kidd, and Cohen (1986), siblings (i.e., DZ twins) should show statistically greater phenotypic comorbidity on average than nonrelated individuals to the extent that they share genes that mediate DRD and ID through pleiotropic effects of these genes, or through the effects of separate genes that are linked. Using same sex DZ twins helps control for the possibility of sex and age effects on DRD or ID that may complicate attempts to detect comorbidity and an etiological relationship. The simplest genetic model for the 11 subtype is one where the same 3 In the Analysis section the logic of the subtype analysis was described. It is worth emphasizing that more complex subtype models could be imagined, and that our failure to find evidence for the subtype we describe does not rule out the existence of other subtypes. 320 GILGER ET AL. genes affect both DRD and ID (pleiotropic effects). (The situation where different, yet tightly linked genes operate to create a 11 subtype is not distinguishable at this level.) The task is to ascertain if the observed DZ concordance for the 11 subtype exceeds what would be expected by chance alone. If there is a genetically mediated 11 subtype, twins selected to be 11 will provide a sample most heavily weighted with this subtype, although some will have both disorders because of chance cooccurrences. Given the available data, the best guess as to the expected DZ chance cooccurrences for DRD and ID, when DRD and ID are independently inherited, are the rates observed for twin pairs ascertained because of having just one of the disorders but not both. By removing all 11 individuals and then taking the cross product of ID and DRD concordances for DZs, an estimate of the 11 concordance rates expected by chance can be obtained. If the observed concordance differs significantly from chance values, then support exists for a genetically based 11 subtype. All tests of concordance and proportion differences are conducted using a z test for proportions (Snedecor & Cochran, 1989; DeFries & Gillis, 1991). For the tests of the general cross-concordance rates between MZs and DZs, we will use a conservative adjustment of the z test where the estimates of p and q in the population are each assumed to be .50. If z approaches significance with p and q equal, we will retest the MZ–DZ difference using the actual MZ and DZ proportions observed to replace the .50 and .50 estimates for p and q. RESULTS Preliminary Analyses To help the reader better assess the meaning of our results pertaining to the associations between DRD and immune problems, we detail several preliminary analyses next. Specifically, we examine the rates observed for AUTO and ALLERGY so that they may be compared to other reports of a similar nature. Next, we present data on the associations between immune disorders and age and sex. Age and sex are correlated with the expression of immune disorders, and these variables should be considered as potential complexities in a study of this sort. We also present information on the degree of positive assortative mating for immune problems. Assortative mating is always an important consideration in family and twin work, the effects of which might modify the estimates of heritability for immune disorders and the ability of our study to identify heritable cosegregation of ID and DRD in our samples. Finally, we summarize data that address the correlations between the immunoassay results with the self or other report immune data. Because other studies of the GBG hypothesis have typically used only survey and, in a few cases, only immunoassay data, there was no way to ascertain 321 TWIN AND FAMILY TABLE 3 Correlations Among Immune System Variables, Sex, and Age for Linkage and Twin Samplesa Linkage c Immune Variable AUTO ALLERGY RA ANA b Twins Twin Subset Sex Age Sex Age Sex Age .13* 2.04 .10 2.09 .28** .03 2.18** 2.01 .00 .05 NA NA .04 .09** NA NA .20* 2.06 .11 .20* .24** .08 2.19 2.07 * p , .05. ** p , .01. a Probands are included in all samples. b See text for details about variables. AUTO 5 composite survey autoimmune score; ALLERGY 5 composite survey allergy score; RA 5 blood serum positivity/negativity for rheumatoid factor; ANA 5 blood serum positivity/negativity for autoimmune antibodies. All variables coded 1 for positive and 0 for negative. c Dummy coded as 1 5 male and 2 5 female. how the two diagnostic methods might compare or modify results. Thus, we also present data that look at the reliability or agreement between the two methods so that this issue can be considered. Rates of Autoimmune and Atopic Disorders. For the AUTO variable, approximately 9.1% of the total linkage sample and 3.0% of the total twin sample were coded positive; the lower rate in the twin sample is likely attributable to the preponderance of children in that sample that have yet to reach the age for risk. For the ALLERGY variable, the rates were quite similar across samples—approximately 27% positivity in the linkage sample and 28% positivity in the twin sample. With the exception of the twin AUTO data, these rates for are slightly higher than those commonly reported by epidemiological studies (e.g., see Table 4; Barrett, 1978; Rich, Fleisher, Schwartz, Shearer, & Strober, 1996). Associations Among Immune Variables and Age and Sex. Pearson correlation coefficients (Howell, 1982) were computed among age, sex, and the immune system function variables for the family and twin samples. The samples included both DRD and non-DRD subjects and probands. Table 3 presents these data and show trends in agreement with expectations based on prior work (Barrett, 1978; Heimer et al., 1963; Mikkelsen et al., 1967; Rich et al., 1996). The correlations in Table 3 show that older subjects in the linkage families were slightly more often positive for RA serum data, and females were somewhat more likely to have an autoimmune-related disorder according to survey reports (for these analyses, females and males are dummy coded as 2 and 1, respectively). Both the overall sample of twins alone as well as the subset of twins and their parents yielded similar patterns 322 GILGER ET AL. of results, although the magnitude of the correlations is somewhat larger for the twin and parent data subset. Assortative Mating for Immune Conditions. Assortative mating can have effects on the outcomes and interpretations of twin analyses (Falconer, 1981). We have previously shown assortative mating to occur for a qualitative diagnosis for DRD as well as a quantitative reading score in families (Pennington et al., 1991; Gilger, 1991), and we were interested if a similar phenomenon may occur for ID. Using ALLERGY or ANA as phenotypes (there was insufficient positivity to examine RA and AUTO), we in fact discovered that 9% of the twin family parents shared ALLERGY positivity, 23% of proband parents of the linkage families shared ALLERGY positivity, and 11% of the linkage parents shared ANA positivity (at a 1/80 or higher dilution). The combination of low rates of ANA and the small number of pairs of twin parents with ANA data, precluded looking for assortion for ANA in the twin families. By taking the square of the population base rates (see bottom of Table 4), the expected rate of assortion, given random mating can be deduced. All the rates of assortative mating for ALLERGY and ANA were greater than chance expectation of 2–4%. Even though we are intrigued, we cannot tell if self or other selection for immune disorders has actually taken place, or whether some sort of shared environmental effect may be responsible for the observed mate commonality for immune problems in these samples. Reliability and Validity of Immunoassay and Questionnaire Data. First, there is the question of how well the AUTO and ALLERGY variables (based on self or other survey reports) correlate with a person’s immunoassay positivity/negativity. Second, there is the related question of the test–retest reliability of the immunoassay data. It is noteworthy that we do not necessarily expect a correlation between ALLERGY and the immunoassay indices, as these measures may tap different physiologic and immunologic systems (Barrett, 1978). However, it is reasonable to examine the agreement between the immunoassay and questionnaire data pertaining to the associations between self-reported autoimmune diagnoses and positivity on the ANA and RA tests in the our sample. Moreover, the interpretation of a lack of agreement between the survey and assay data is ambiguous: either or both indices may be unreliable or invalid, or the two indices may tap reliable and valid immune functions that are different. Nonetheless, we present these data for the reader’s interest. There was not a significant association between AUTO and either ANA or RA. The percent of subjects who were immunoassay ANA positive was very similar in the groups with (21%) and without (20%) self-reported autoimmune illnesses. For RA positivity, the rates were likewise similar across the two groups, 17 and 16%, respectively. For a subset of the linkage sample, immunoassays were performed twice, TWIN AND FAMILY 323 both within and across laboratories to evaluate the test–retest and cross-lab reliability of these assays. (The second laboratory was coauthor Kotzin’s.) The reliability across labs was assessed by assaying the same blood samples in both labs. For the ANA assay on 119 subject samples, the overall agreement was 74% at slightly different dilutions in each lab (1/64 and 1/80), with 16% of the samples being positive in both labs. For the RA assay, the agreement for 102 subject samples was 63%, with 4.9% being positive in both labs. The within-lab reliability was assessed by blindly assaying two separate blood samples from the same individuals (N 5 14). For the ANA assay, the agreement across samples was 100%; for the RA assay it was 90%. In sum, there was excellent within-lab reliability, and weaker cross-lab reliability. Oversensitivity of the assays, particularly ANA, was found in both labs, consistent with the literature cited earlier. Association between DRD and ID4 Survey Data. Table 4 shows that the AUTO rates in the twins are near zero or very close to the population base rates. This would be expected given that most of the twins are not yet at the age of onset typical for many autoimmune illnesses (see footnote 2). However, the self report for autoimmune disorders does increase in the twin parents and ranges from 6 to 21%, depending on the subsample being examined. The only significant difference found by a z test between proportions was for the AUTO rates of the subset of 33 parents from the DRD proband twin group and the population expectancy of 4.7% (p , .01). Because the AUTO rates in the twin subset are much higher than in the original twin sample, there is some suggestion that the subset sample may suffer from a self-selection artifact. Table 4 also shows no evidence of a significantly greater rate of either autoimmune or allergic disorders in DRD individuals (10–11% positivity) compared to non-DRD individuals (10–39% positivity) in the linkage families. In fact, the expected direction of the trend was reversed, with higher ALLERGY rates in the non-DRD subjects (p , .01). However, 42% of the 19 DRD linkage family probands were positive for ALLERGY. When compared to the ALLERGY rates in DRD relatives, the higher rates in the proband sample again suggests the potential effects of biased ascertainment. Finally, more married-ins of the linkage sample were ALLERGY positive than would be expected by chance alone (31% versus 2–4%; see earlier discussion of assortative mating). For the linkage sample, rates of reported autoimmune illness were 11% 4 Perhaps because the amount of variance explained by sex or age in either the AUTO or ALLERGY variables is quite small, the conclusions suggested by our main analyses were not changed when sex or age was controlled for, and we will not present the results of these analyses here. 324 GILGER ET AL. TABLE 4 Rates of Autoimmune and Allergic Disorders According to Immunoassay and Survey Dataa Sample Twin Family Sample DRD proband group DRD N Non-RD N Non-DRD proband group DRD N Non-RD N Parents of Twin Pairs DRD proband parents N Non-DRD proband parents N Twin Family Subsample DRD proband pairs N Non-DRD proband pairs N Parents of Twin Pairs DRD proband parents N Non-DRD proband parents N Linkage Families DRD relatives N Non-DRD N Proband N Married-ins Population Base Rates d AUTO ALLERGY ANA RA ,.01 549 ,.01 295 .23 b,c 549 .32 c 295 NA NA NA NA 0 28 .01 466 .21 23 .28 c 466 NA NA NA NA .29 c NA NA .31 c NA NA .06 531 .07 c 314 531 314 0 40 .04 24 .17 40 .29 c 24 .21 c 33 .14 22 .39 c 33 .14 22 .11 c 135 .10 c 94 .10 19 .05 57 2–5% .21 39 .33 24 .25 32 .41 22 .10 b 135 .39 c 94 .42 c 19 .31 c 57 13–22% .17 99 .21 78 .14 14 .25 44 NA .04 24 .09 11 .21 15 .09 11 .11 62 .15 41 .28 7 .07 27 NA a See text for details about variables. AUTO 5 composite survey autoimmune score; ALLERGY 5 composite survey allergy score; RA 5 blood serum positivity/negativity for rheumatoid factor; ANA 5 blood serum positivity/negativity for autoimmune antibodies. b Significant z test for proportion difference between DRD and non-DRD relatives, p , .05. All tests are one-tailed unless noted otherwise in text. c Significant z test for proportion difference between DRD, non-DRD relatives and population base rates (p , .05). Range of population rates are shown in this table, and specific values of 4.7 and 13% were used for AUTO and ALLERGY comparisons, respectively. All tests are one-tailed unless noted otherwise in text. d Estimates only. Based on summary rates in Barrett (1978), Gilger et al. (1992), and Rich et al. (1996). TWIN AND FAMILY 325 for DRD relatives, 10% for non-DRD relatives, 10% for the probands, and 5% for the married-ins. The DRD rate was significantly higher than the population base rate of 4.7% (p , .01), and the rate of positivity in the non-DRD relatives approached significance compared to the population base rate (p , .10, two-tailed). Other significant differences were found when comparisons were made to population base rates, and these are noted in Table 4. In sum, the survey data in Table 4 do not suggest a robust association between DRD and AUTO or ALLERGY variables. This is particularly true when we make the essential comparison between DRD and non-DRD relatives, where certain ascertainment biases can be controlled. Even though occasional significant results were obtained when comparisons were made to population base rates, these differences were often unexpected (i.e., higher rates in non-DRD individuals). It is also noteworthy that the linkage probands and the twin subset have higher rates of immune disorders. This further highlights the need for careful attention to ascertainment biases in these types of studies. In general, these results are similar to one earlier report of ours on four large family studies (two of which were portions of the linkage pedigrees and the twin families reported herein; Gilger, Pennington, & DeFries, 1992), but contrary to an even earlier report that looked at some preliminary data at the initial stages of the linkage family study (Pennington, Smith, Kimberling, Green, & Haith, 1987). Immunoassay Data. The immunoassay positivity rates are also shown in Table 4. Rates of ANA and RA positivity were not significantly elevated in DRD compared to non-DRD relatives. In this sense, conclusions based on the immunoassay data are essentially consistent with those based on the survey data. Also consistent with the survey data are the results in the linkage families, where slightly higher immunoassay positivity rates are found in the married-ins and non-DRD blood relatives. The ANA and RA rates in the twin family subset are similar in pattern, although generally higher. Because of the potential oversensitivity of the assays and problems with establishing reliable population base rates for assay data, we do not provide statistical tests between the rates we observe and the population expectancies as we did for the survey data. We also examined whether there were group differences at higher dilutions to diagnose positivity. This helps control for the possibility of confounding effects of oversensitive assay tests. Using a cutoff of 1/160 for the ANA test, in the linkage sample we found positivity for 0% of the probands, 8.1% of the DRD relatives, 12.8% of the non-DRD relatives, and 11.4% of the married-ins. In the twin subset sample, rates of ANA positivity at 1/160 or higher were 12.8% for DRD twins, 7.1% for their parents, 20.8% for nonDRD twins, and 22.7% for their parents. If anything, rates at this higher cutoff were greater in non-DRD individuals than in DRD individuals. Thus, the null results (i.e., no association between DRD and positivity) in Table 4 are not simply an artifact of an oversensitive cutoff. 326 GILGER ET AL. Tests of a Common Etiology Between DRD and ID Before conducting cross-concordance analyses or tests for a genetic correlation between DRD and ID, we would typically require a correlation between DRD and ID in our twin sample. In theory, we would expect phenotypic associations for two traits when they share a common etiology. However, it is possible that some genetic relationship exists that is somehow obscured due to etiologic heterogeneity, variable expressivity, and incomplete penetrance. The subtype analysis presented in this section addresses this phenomenon, although we present the general cross-concordance analysis as well (see Analyses section previously). Independent Heritability of DRD and ID. Qualitative diagnoses for DRD and ALLERGY were found to be heritable in the twin sample. Specifically, the probandwise concordance for DRD in MZs was .85 and in DZs it was .61 (z 5 4.18, p , .01). For ALLERGY, the MZ probandwise concordance was .83 which was significantly larger than the DZ rate of .49 (z 5 5.15, p , .01). Genetic Correlation Between DRD and ID. General cross-concordances were computed for DRD and ALLERGY. First, when selecting for a DRD proband and then looking for ALLERGY in the cotwin, the MZ rate was .21 and the DZ rate was .28 (z 5 21.17, p . .05). When looking first for an ALLERGY positive proband, then a DRD cotwin, the rates were .34 for MZs and .42 for DZs (z 5 21.04, p . .05). Thus, there is no significant evidence for a genetic correlation between DRD and ALLERGY. This finding agrees with the data from several family data sets and a smaller data set of twins reported earlier (Gilger, Pennington, Green et al., 1992). Test for a Heritable Subtype. The analyses performed thus far do not show the expected association between DRD and ID nor any other general common genetic etiology. However, there may be a subgroup or a subtype of DRD that shares a common biologic etiology with ID and that is being ‘‘masked’’ by heterogeneity in our data set. Therefore, we conducted a specialized form of subtype concordance analyses (see Analysis section previously). For the DRD–ALLERGY subtype, all DZ twin pairs where at least one member was 11 were selected, and the rates of the 11 subtype in the cotwins was examined relative to expectations given the independent prevalences and heritabilities of the two disorders. The observed 11 concordance rate for DZ twins was .15. This was not significantly larger than the expected chance 11 concordance of .25 (p . .05), and in fact it was in a direction opposite of expectation if there was a genetic basis to a DRD–ID subtype.5 5 We use a slightly different method to look for a heritable component to comorbidities in twins in Gilger, Pennington, and DeFries, 1992 and Gilger, Pennington, Green et al., 1992. Originally, this method included an MZ–DZ comparison, although, technically, all that is required is a DZ comparison to chance expectations as was done here. Even though we ap- TWIN AND FAMILY 327 DISCUSSION We conducted a study of the association between DRD and ID using both survey and immunoassay data in two separate samples. We failed to find an association between DRD and ID in individuals or families using either the survey or immunoassay data. Tests for heritability of both DRD and allergic conditions were significant in the twin sample, but there was no evidence of a genetic correlation between the two disorders—in the sample as a whole or as a subtype. Although we found some weak evidence for an association between DRD and ID in an earlier study of a subset of linkage families (e.g., Pennington et al., 1987), increasing the sample size has eliminated the association. These results join other null results of previous works (Bryden et al., 1994). The present results and the results of other family studies presented in Gilger, Pennington, Green et al. (1992) also fail to support the hypothesis that the QTL on chromosome 6 shown to influence DRD (Cardon et al., 1994; Grigorenko et al., 1996) exerts a pleiotropic effect on the forms of IDs examined in this paper. This pleiotropic effect seemed possible given that the QTL for DRD lies somewhere in the HLA region. Thus, our data suggest that if there is a subtype of DRD that is associated etiologically with ID, it is likely to be different from the DRD linked to chromosome 6, or it may be restricted to a very small subset of the DRD population that was not represented to any great extent in the two separate DRD samples we examined here. Unfortunately, we did not have sufficient data to examine the correlations among DRD and subtypes of ID. Not all the immune disorders we or other authors have assessed reflect the effects of major genes on chromosome 6 or in the HLA region. It remains possible that the GBG hypothesized associations may be more robust in more homogeneous samples of DRD subjects where HLA-and non-HLA-based IDs can be distinguished (e.g., Bias et al., 1986; Sawcer, Jones, Feakes, Gray, Smaldon, Chataway, Robertson, Clayton, Goodfellow, & Compston 1996). In our discussion of the Bryden et al. (1994) paper, we questioned the utility of looking for trait associations/correlations as predicted by the Geschwind–Behan–Galaburda model to help us understand the etiology of DRD and ID (Gilger & Pennington, 1995). The phenotypic associations proposed by the GBG model would be produced by a long chain of complex physiologic events which may or may not make sense biologically (Bryden et al., 1994). However, even if these associations are found, they tell us little about an underlying common etiology for DRD and ID. The associations predicted by the GBG model can be thought of as examproach comorbidity differently in this earlier work, looking at just the DZ rates yields essentially the same results: weak support for a 11 subtype for DRD–attention deficit disorder in Gilger, Pennington, and DeFries (1992) and no support for a DRD–ID subtype in Gilger, Pennington, Green et al. (1992). 328 GILGER ET AL. ples of comorbidity, where a person is afflicted with more than one disorder, such as developmental reading disability and some form of immune system dysfunction. Logically, comorbidity for ID and DRD (or any associated traits) beyond chance expectations can be thought of as reflecting some or all the following mechanisms (Gilger & Pennington, 1995; Caron & Rutter, 1991): (1) both ID and DRD are a result of the same basic biological/developmental process (i.e., the common etiology explanation); (2) both ID and DRD are predominately etiologically independent disorders, even though there is a subgroup of DRD individuals that have comorbid ID because of a common etiology (i.e., the subtype explanation); (3) one of the disorders, say DRD, is caused by the other disorder, ID, in a fashion that mimics the form of DRD that is caused by an independent biological/developmental factor (i.e., the phenocopy explanation); and/or (4) the cooccurrence of DRD and ID is largely the result of chance or systematic errors in methodology (i.e., the artifact explanation). Given that both DRD and ID have been shown to be independently heritable, it makes sense to think of these four mechanisms in terms of the presence or absence of a genetic basis. In this context, the common etiology explanation would suggest that the same or tightly linked genes are responsible for a significant proportion of the covariance for DRD and ID. The subtype explanation suggests essentially the same thing, but only for a subset of all DRD individuals. In other words, for most individuals with DRD, their disorder may be a function of biologic factors unrelated to ID, whereas others have a form of DRD that is biologically related to ID. Such a situation, if it did exist, would help explain the inconsistent findings of previous work on the GBG hypothesis. The phenocopy explanation suggests DRD and ID may be associated in some people because having one disorder can ‘‘trigger’’ the other. For example, the stress of having DRD may change immune functions and produce ID as a secondary consequence. And finally, the artifact explanation emphasizes the genetic independence of the two disorders and how selection biases could artificially inflate the comorbidity rates observed. The Bryden et al. (1994) summary, as well as our own research, suggests that the positive reports for DRD–ID associations are most consistent with options two and four, with little evidence that a large proportion of the DRD– ID covariance is the result of a direct common genetic etiology. The likelihood of the phenocopy explanation is very low and can probably be dismissed as well. In large part Bryden et al. (1994) suggest similar conclusions because the associations across studies they review are weak and inconsistent. That some of the previous reports showing an association between ID and DRD may reflect sampling bias is also supported by our data (Bryden et al., 1994). Rates of AUTO and ALLERGY were elevated in the DRD probands of the linkage families compared to their relatives. These rates were also raised in the subset of twin family volunteers compared to the overall twin TWIN AND FAMILY 329 sample that was ascertained via an epidemiologic scheme. We also showed this ascertainment effect in two other family samples presented in Gilger, Pennington, Green et al. (1992). Thus, prior work on the GBG model that has relied on proband and volunteer samples may be of sufficient bias to be nonrepresentative of the general DRD population. There are several caveats to our conclusions. One is the problem of survey and/or immunoassay reliability, and the related issue of how age effects on risk for certain immune disorders may have led to missed or incorrect identification of DRD children carrying the biotype for later developing ID (see also footnote 2). Setting aside for the moment the potential unreliability of the survey data and assay data, we are then faced with the noticeable trend for moderately high rates of atopias, and to a lessor extent autoimmune disorders, to exist in our families. Thus, even though ID and DRD were not correlated, bot DRD and ID were somewhat familial. This is expected for traits with significant heritability, but why should atopias be elevated in families selected through DRD probands relative to population base rates? This familial tendency is perhaps predicted by the GBG model even though there is no evidence for a DRD–ID etiologic link. What then would explain this phenomenon? Three possibilities we think most worthy of consideration are as follows. First, there may be some sort of maternal effect on the fetus that modifies the risk for allergies in offspring (e.g., fetal–maternal histoincompatibility; Gualtieri & Hicks, 1985; Crawford, Kaplan, & Kinsbourne, 1992, 1994; Lahita, 1988), although cross-fostering studies in mice with immune dysfunction do not support a maternal effect (see Galaburda, Schrott, Sherman, Rosen, & Dennberg, 1996). On the other hand, Gilger, Pennington, Green et al. (1992) report significantly higher rates of miscarriages in Iowa families of DRD children compared to controls, and this is in some ways consistent with a maternal immunological attack on the fetus. Second, our data suggest that there is a familial factor that is shared among relatives that tends to raise their likelihood of having (or at least reporting) an allergic condition. Twin and pedigree data show larger than expected correlations among relatives for allergies. Although some proportion of this covariance can be explained by shared genes, the higher than normal population rates of ID cannot be explained given the manner in which the samples were ascertained (i.e., through a DRD proband). This, along with the significant assortative mating for allergies in our samples, suggests that there may be a within-family environmental trigger that is in part responsible for the allergies common to relatives. This effect could be historical in nature: even before the GBG hypothesis was fully described, the lay literature and folklore contained descriptions of various immune and other traits supposedly associated with DRD. Such historical information could lead to the overreporting or greater self-awareness of IDs in families with DRD relatives. However, while self and other report bias may have contributed to the high rates of 330 GILGER ET AL. ID in our samples, such biases cannot account for the similarly high rates of immunoassay positivity we observed. Finally, because the majority of our subjects resided in the same general geographic location (Colorado and surrounding areas), it is possible that some area-specific allergens elevated the rate of IDs, and thereby masked special DRD–ID associations resulting from a common genetic etiology. Precisely how such an effect would be manifested, particularly in light of the elevated rates we found in both DRD and non-DRD subjects is difficult to predict, although it should be considered in future work. In conclusion, we provide evidence that fails to support the GBG model that postulated a common biologic etiology for DRD and ID. As Bryden et al. (1994) point out, the GBG model has had a very long history and has been explored by literally hundreds of researchers and studies, and negative findings of well-designed studies are more common than positive findings (see also, Gilger et al., 1992), and most of the studies of the GBG model have been small scale, dealt with clinically ascertained populations, and used only survey methodologies. Even though our results are negative, it is still possible that the long chain of physiologic events described by the GBG model are real but that they yield only very small and occasional associations between ID and DRD. Then we are left with the question of the practical and research utility of a model with such low predictive power. Whatever the answer to this question, continuing to repeat small-scale, clinical-drawer studies is not likely to advance the field of learning disabilities and neuropsychology, and they will not address the truly interesting questions raised by the GBG model. Thus, although we do not advocate completely abandoning the GBG hypothesis about developmental disabilities, we do suggest studying the model with different methodologies and from different perspectives. For example, some animal models may later suggest specific genes for ID and brain anomalies that we could later test in DRD individuals (e.g., Galaburda et al., 1996). It is also possible that the hypothesized relationship between ID and DRD needs modification and can be more appropriately tested by later research. In sum, the future of the GBG model rests with new studies that do more than simply test for the proposed associations at the phenotypic level. Genetic work and animal models testing neurobiological mechanisms will likely lead to a clarification of the strengths and weaknesses of the GBG model in terms of the etiology of developmental disorders. For an example of candidate gene linkage work suggested by a mouse model or DRD and JD, see Gilger, Pennington, Smith, Harbeck, and Kotzin (1995). REFERENCES Barrett, J. T. 1978. Textbook of immunology: An introduction to immunochemistry and immunobiology. St. Louis: Mosby. TWIN AND FAMILY 331 Bias, W. B., Reveille, J. D., Beaty, T. H., Meyers, D. A., & Arnett, F. C. 1986. Evidence that autoimmunity in man is a mendelian dominant trait. American Journal of Human Genetics, 39, 584–602. Bryden, M. P., McManus, I. C., & Bulman-Fleming, M. B. 1994. Evaluating the empirical support for the Geschwind–Behan–Galaburda model of cerebral lateralization. Brain and Cognition, 26, 103–167. Brain and Cognition, 26,(2) 1994, 27 (1), 1995. Cardon. L., Smith, S. D., Fulker, D. W., Kimberling, W. J., Pennington, B., & DeFries, J. C. 1994. Quantitative trait locus for reading disability on chromosome 6. Science, 266, 276– 279. Caron, C., & Rutter, M. 1991. Comorbidity in child psychopathology: Concepts, issues, and research strategies. Journal of Child Psychology, Psychiatry, and Allied Disciplines, 32, 1063–1080. Cook, L., & Agnello, V. 1992. Tests for detection of rheumatoid factors. In N. R. Rose, E. C. DeMacario, J. L. Fahey, H. Friedman, and G. M. Penn (Eds.), Manual of clinical laboratory immunology (4th ed. pp. 762–764). Washington, DC: American Society for Microbiology. Crawford, S. G., Kaplan, B. J., & Kinsbourne, M. 1992. The effects of parental immunoreactivity on pregnancy, birth, and cognitive development: Maternal immune attack on the fetus? Cortex, 28, 483–491. Crawford, S. G., Kaplan, B. J., & Kinsbourne, M. 1994. Are families of children with reading difficulties at risk for immune disorders and nonrighthandedness? Cortex, 30, 281–292. DeFries, J. C. 1991. Gender ratios in children with reading disability and their affected relatives: A commentary. Journal of Learning Disabilities, 22, 544–545. DeFries, J. C., Filipek, P. A., Fulker, D. W., Olson, R. K., Pennington, B. F., Smith, S. D., & Wise, B. W. 1997. Colorado Learning Disabilities Research Center. Learning Disabilities Quarterly: A Multidisciplinary Journal, 8, 7–19. DeFries, J. C., & Gillis, J. 1991. Etiology of reading deficits in learning disabilities: Quantitative genetic analysis. In J. E. Obrzut & G. W. Hynd (Eds.), Advances in the neuropsychology of learning disabilities: Issues, methods, & practice (pp. 29–47). Orlando, FL: Academic Press. Dunn, L. M., & Markwardt, F. C., Jr. 1970. Manual for the Peabody Individual Achievement Test. Circle Pine, MN: American Guidance Service. Falconer, D. S. 1981. Introduction to quantitative genetics (2nd ed.). London: Longman House. Fielder, A. H. L., Walport, M. J., Batchelor, J. R., Rynes, R. I., Black, C. M., Dodi, I. A., & Hughes, G. R. V. 1983. Family study of the major histocompatibility complex in patients with systemic lupus erythematosus: Importance of null alleles of C4A and C4B in determining disease susceptibility. British Medical Journal, 286, 425–428. Fletcher, J. M. 1992. The validity of distinguishing children with language and learning disabilities according to discrepancies with IQ: Introduction to the special series. Journal of Learning Disabilities, 25(9), 546–548. Fritzler, M. J. 1992. Immunofluorescent antinuclear antibody test. In N. R. Rose, E. C. DeMacario, J. L. Fahey, H. Friedman, and G. M. Penn (Eds.), Manual of clinical laboratory immunology (4th ed., pp. 724–729). Washington, DC: American Society for Microbiology. Galaburda, A. M., Schrott, L., Sherman, G. F., Rosen, G. D., & Denenberg, V. H. 1996. Animal models of developmental dyslexia. In C. Chase, G. Rosen, & G. Sherman (Eds.), Developmental dyslexia: Neural, cognitive, and genetic mechanisms (pp. 3–14). York Press: Baltimore, MD. 332 GILGER ET AL. Geschwind, N., & Behan, P. 1982. Left-handedness: Associations with immune disease, migraine, and developmental learning disorder. Proceedings of the National Academy of Science, USA, 79, 5097–5100. Geschwind, N., & Behan, P. O. 1984. Laterallity, hormones, and immunity, In N. Geschwind & A. Galaburda (Eds.), Cerebral dominance: The biological foundations. Cambridge, MA: Harvard University Press. Geschwind, N., & Galaburda, A. M. 1985a. Cerebral lateralization: Biological mechanisms, associations, and pathology: I. A hypothesis and a program of research. Archives of Neurology, 42, 428–459. Geschwind, N., & Galaburda, A. M. 1985b. Cerebral lateralization: Biological mechanisms, associations, and pathology: II. A hypothesis and a program of research. Archives of Neurology, 42, 521–552. Gilger, J. W. 1991. Differential assortative mating found for academic and demographic variables as a function of time of assessment. Behavior Genetics, 21(2), 131–150. Gilger, J. W., Hanebuth, E., Smith, S. D., & Pennington, B. F. 1996. Differential risk for developmental reading disorders in the offspring of compensated versus noncompensated parents. Reading and Writing: An Interdisciplinary Journal, 8, 407–417. Gilger, J. W., & Pennington, B. F. 1995. Why associations among traits do not necessarily indicate their common etiology: A comment on the Geschwind–Behan–Galaburda model. Brain and Cognition, 27, 89–93. Gilger, J. W., Pennington, B. F., & DeFries, J. C. 1992. A twin study of the etiology of comorbidity: Attention deficit hyperactivity disorder and dyslexia. Journal of the American Academy of Child and Adolescent Psychiatry, 31(2), 343–349. Gilger, J. W., Pennington, B. F., Green, P., Smith, S. A., & Smith, S. M. 1992. Dyslexia, immune disorders, and left-handedness: Twin and family studies of their relations. Neuropsychologia, 30(3), 209–227. Grigorenko, E. L., Wood, F. B., Meyer, M. S., Hart, L. A., Speed, W. C., Shuster, A., & Pauls, D. L. 1996. Susceptibility loci for distinct components of developmental dyslexia on chromosomes 6 and 15. American Journal of Human Genetics, 60, 27–39. Gualtieri, T., & Hicks, R. E. 1985. An immunoreactive theory of selective male affliction. Behavioral and Brain Sciences, 8, 427–441. Hannah, M. C., Hopper, J. L., & Mathews, J. D. 1983. Twin concordance for a binary trait: I. Statistical methods illustrated with data on drinking status. Acta Geneticae Medicae et Gemellologiae, 32, 127–137. Heimer, R., Levin, F. M., & Rudd, E. 1963. Globulins resembling rheumatoid factor in serum of the aged. American Journal of Medicine, 35, 175–181. Hopper, J. L., Hannah, M. C., Macaskill, G. T., & Mathews, J. D. 1990. Twin concordance for a binary trait: III. A bivariate analysis of hay fever and asthma. Genetic Epidemiology, 7, 277–289. Howell, D. C. 1982. Statistical methods for psychology. Boston: Duxbury Press. Lahita, R. G. 1988. Systemic lupus erythematosis: Learning disability in the male offspring of female patients and relationship to laterallity. Psychoneuroendocrinology, 13, 385– 396. Libscomb, M. R., Cope, L. D., Stephens, G. L., Deng, J. S., & Gilliam, J. N. 1984. Comparison of substrates for the detection of antinuclear antibodies in normals and in patients with connective tissue and other diseases. Diagnostic Immunology, 2, 181–187. Lykken, D. T., Tellegen, A., & DeRubeis, R. 1978. Volunteer bias in twin research: The rule of two thirds. Socio Biology, 25, 1–9. Mikkelsen, W. M., Dodge, H. J., Duff, I. F., & Kato, H. 1967. Estimates of the prevalence TWIN AND FAMILY 333 of rheumatic diseases in the population of Tecumseh, Michigan, 1959–1960. Journal of Chronic Diseases, 20, 351–369. Neale, M. C., & Cardon, L. R. 1992. Methodology for genetic studies of twins and families. London: Kluwer Academic Press. Nichols, R. C., & Bilbro, W. C. 1966. The diagnosis of twin zygosity. Acta Geneticae Medicae et Gemellologiae, 16, 265–275. Pauls, D. L., Hurst, C. R., Kruger, S. D., Leckman, S. D., Kidd, K., & Cohen, D. J. 1986. Gilles de la Tourette’s syndrome and attention deficit disorder with hyperactivity: Evidence against a genetic relationship. Archives of General Psychiatry, 43, 1177–1179. Pennington, B. F., Gilger, J. W., Olson, R., & DeFries, J. 1992. External validity of age versus IQ discrepant diagnoses in reading disability: Lessons from a twin study. Journal of Learning Disabilities, 25 (special), 562–573. Pennington, B. F., Gilger, J. W., Pauls, D., Smith, S. A., Smith, S., & DeFries, J. C. 1991. Evidence for major gene transmission of developmental dyslexia. Journal of the American Medical Association, 266(11), 1527–1534. Pennington, B. F., Smith, S. D., Kimberling, W. J., Green, P. A., & Haith, M. M. 1987. Lefthandedness and immune disorders in familial dyslexics. Archives of Neurology, 44, 634– 639. Rich, R. R., Fleisher, T. A., Schwartz, B. D., Shearer, W. T., & Strober, W. 1996. Clinical immunology: Principles and practice. St. Louis: Mosby. Sawcer, S., Jones, H. B., Feakes, R., Gray, J., Smaldon, N., Chataway, J., Robertson, N., Clayton, D., Goodfellow, P. N., & Compston, A. 1996. A genome screen in multiple sclerosis reveals susceptibility loci on chromosome 6p21 and 17q22. Nature Genetics, 13, 464–468. Shaywitz, B. B., Shaywitz, S., Fletcher, J., and Escobar, M. 1990. Prevalence of reading disability in boys and girls. Journal of the American Medical Association, 264, 998–1002. Smith, S. D., Kimberling, W. J., Pennington, B. F., & Lubs, H. 1983. Specific reading disability: Identification of an inherited form through linkage analysis. Science, 219, 1345– 1347. Snedecor, G. W., & Cochran, W. G. 1989. Statistical methods. Ames, IA: Iowa State Press. Sontheimer, R. D. 1993. A young woman with a photosensitive butterfly facial rash, fatigue, anemia, and positive antinuclear antibody. Arthritis and Rheumatism, 36, 871–874. Urion, D. K. 1988. Nondextrality and autoimmune disorders among relatives of learningdisabled boys. Annals of Neurology, 24, 267–269. Wechsler, D. 1974. Examiner’s manual: Wechsler Intelligence Scale for Children—Revised. New York: The Psychological Corporation. Wechsler, D. 1981: Examiner’s manual: Wechsler Intelligence Adult Scale—Revised. New York: The Psychological Corporation.