Panic and Phobic Anxiety: Phenotypes, Endophenotypes, and Genotypes

In an era dominated by psychogenic theory, the reports of a substantially higher concordance for schizophrenia in identical than fraternal twins caused consternation, at least in those who noticed. Such persuasive findings were not definitive, since alternative hypotheses concerning blurred ego boundaries, differential familial environment, etc., were raised.

The success of identifying the variations in DNA sequence that cause disorders such as Huntington’s disease (autosomal dominant) have not as yet proven definitive for psychiatric diseases, perhaps because of their more obscure familial patterns. In this issue of the Journal, Smoller and Tsuang argue that even a disease determined entirely by a single gene still presents a major problem, since similar-looking diseases may stem from quite different genetic sources. One group of geneticists may report a linkage with one particular chromosomal location, whereas another distant group of geneticists, nominally pursuing the same diagnosis, may find linkage to another chromosomal location. Of course, this has happened, and they may both be right.

Smoller and Tsuang state that the categorical phenotypes of our diagnostic system are of genetic relevance since they demonstrate familiality and heritability, but they nonetheless may not be the nosology most relevant to identifying genetic loci, given the problems of phenocopies, variable expressivity, and incomplete penetrance. In a previous paper Tsuang et al. (1) illuminated the problem of categorical phenotypic error.

Smoller and Tsuang carefully review familial and twin studies of the anxiety disorders, with particular regard to panic disorder, agreeing that familial evidence indicates that panic and phobic disorders tend to “breed true” but that areas of ambiguity remain.

They state, “For the purposes of identifying susceptibility genes, the practice of condensing phenotypic data into dichotomous or categorical diagnoses results in a loss of information. . . .” However, that is true only if an underlying continuum is being arbitrarily segmented. Several studies have expanded the affected state’s definition beyond strict criteria to include subsyndromal variants, which has improved the genetic analysis. However, is that loss of information due to categorization or simply drawing the category’s boundaries too tightly?

DSM-IV has practical, clinical purposes. The criteria not only reliably discriminate diagnoses from each other but also allow the inference that there is sufficient disability to justify the sick role. The patients are not only diseased, they are ill (2). With advanced imaging and biochemical techniques, it is now common to detect disease in asymptomatic, manifestly unimpaired patients. From the point of view of linkage genetics, these diseased but not ill patients are false negatives.

Descriptive attempts to discriminate minor manifestations of disease from similar aspects of normal functioning may be impossible. To avoid the Scylla of false negatives, one may fall into the Charybdis of false positives. Current mathematical approaches to linkage studies are more damaged by false positives than false negatives, which seems to argue for tight boundaries.

Smoller and Tsuang suggest that the arbitrariness of categorical diagnoses could be minimized by viewing anxiety phenotypes as dimensions. To support this, they cite evidence for a “general neurotic syndrome.” This seems irreconcilable with the data for the anxiety disorders that breed true, so this is difficult to follow. They cite Australian studies indicating substantial genetic influence on neuroticism and, to a lesser degree, on symptoms of anxiety and depression, but not the heritability of categorical anxiety disorders.

How does moving toward dimensionality help resolve this straight-out factual contradiction? Further, it is ambiguous to refer to “quantitative traits such as shyness.” The measure may be a quantitative scale, but the phenomenon’s nature is the issue. For instance, Kagan considers behavioral inhibition as due to a qualitative disjunction despite his use of a quantitative scale (3).

Does a dimensionality emphasis help? To optimize the definition of “genetic cases,” Smoller and Tsuang suggest that we focus on phenotypic extremes with an early onset. That seems sensible but irrelevant to dimensionality. They also suggest analyzing family and twin data to determine which phenotypic features are associated with the greatest familiality and heritability, but this falls afoul of the same problem: multiple phenocopies. The authors even suggest that one should forswear a priori hypotheses and simply analyze the range of phenotypic data in an attempt to predict caseness.

This sounds uncomfortably close to attempts to eke out, by multiple regression techniques, patient subtypes that respond best to particular pharmacological interventions. Such efforts regularly failed, in part because of excessive noise in the data. However, specific diagnostic categorical hypotheses, e.g., atypical depression and obsessive-compulsive disorder, paid off in terms of psychopharmacological specificity.

Smoller and Tsuang also suggest that investigators may discover symptom clusters with independent genetic influences, referring to prominent respiratory symptoms in panic disorder. They note that Briggs et al. (4), using factor and cluster analysis, distinguished patients with dyspneic panic attacks, who responded better to imipramine than alprazolam. The patients with nondyspneic panic attacks responded better to alprazolam than imipramine. Further, Perna et al. (5) found that subjects with a history of unexpected panic attacks had a high rate of family history of panic disorder and that first-degree asymptomatic relatives of panic disorder patients had a much higher rate of carbon dioxide sensitivity than normal subjects (6). Further, Perna et al. (7) showed that the panic disorder probands with CO₂ hypersensitivity accounted for most of the familial loading. CO₂ hypersensitivity may be due to a particular genetic dysfunction among the multiple phenotypes called panic disorder, and it may even cut across current nosological boundaries.

Smoller and Tsuang argue that “by examining temperaments and traits that may underlie anxiety disorders, we may circumvent the problem of the ambiguous boundaries of diagnostic categories.” Yet all examples given are of extreme cases, which yield higher interrater reliability, less diagnostic error, and more homogeneity, without any necessary assumption that the phenomenon is best seen as a continuum.

The authors close with a question: “Whether anxiety phenotypes can be collapsed with depression into a ‘general neurotic syndrome’ . . . .” It seems to me that the weight of evidence with regard to breeding true, in these authors’ summary, make this hypothesis untenable.

In the field of experimental therapeutic comparisons, the overriding finding has been a strong allegiance effect. Those who believe that a certain treatment is the best are just the investigators most likely to find that this is so. Similarly, in the area of familial study, those with an a priori hypothesis of a general neurotic syndrome seem to find supporting data, whereas we “splitters” find that diseases breed true. In experimental therapeutics, a distinct advance is the active collaboration of experts with opposing theoretical presumptions (unpublished works by M. Liebowitz et al. and D. Barlow et al.). Perhaps the same will prove necessary for genetics.

The relevance of respiratory CO₂ sensitivity to the genetics of panic disorder receives remarkable confirmation in this issue by Bellodi et al., who amplify the classic diagnostic concordance study of identical and fraternal twins by administering CO₂ challenges. With regard to panic disorder, probandwise concordance rates were higher for monozygotic pairs (six of nine, 67%) than for dizygotic pairs (neither of two, 0%). For spontaneous panic attacks, the respective rates were 71% and 18%. For CO₂-induced panic attacks, the respective rates were 56% and 13%. These marked differences, if replicated in larger samples, indicate that the genetic relationship is not simply additive. Such complex genetics make even more problematic attempts to link disease to single DNA regions.

Also in this issue, Bisaga et al. present a follow-up of positron emission tomography (PET) studies by Reiman et al., who used an ¹⁵O-labeled tracer and reported that patients with panic disorder who were vulnerable to lactate at baseline had abnormally greater right than left parahippocampal blood flow, blood volume, and oxygen metabolism, as well as an abnormally high whole-brain metabolism.

Using PET with [¹⁸F]fluorodeoxyglucose, Bisaga et al. found significant differences in glucose metabolism. However, it was the left (rather than right) parahippo­campal areas of these panic disorder subjects that had high rates of glucose metabolism. There also was low glucose metabolism in the right inferior parietal and right superior temporal brain regions.

The search for cerebral markers is of great interest, but lacking a detailed theory of how psychopathology relates to cerebral dysfunction, we must recognize that this is useful, rather than definitive, exploratory work. Unfortunately, the history of biological psychiatry is replete with reports of baseline differences between patients and normal subjects that turn out to be artifacts, since these are naturalistic, multiply confounded, rather than experimental, studies. I believe the differences between patients and normal comparison subjects that have stood up best have been the outcome of challenge studies, e.g., lactate infusion, sedation threshold, CO₂ inhalation. Much psychopathology may be due to adaptive deficiencies in cybernetic control mechanisms, best revealed by perturbing the system rather than simply observing it at rest. Combining challenges with genetic studies may prove a useful strategy in dealing with the multiple phenocopy problem.

Address reprint requests to Dr. Klein, Unit 22, 722 West 168th St., New York, NY 10032; [email protected]. Supported in part by grant MH-30906 from NIMH.