In this issue of the Journal, Grillon et al. (1) report a possible psychophysiological marker of panic disorder. In their study, participants with a diagnosis of panic disorder and healthy comparison subjects were presented with colored geometric shapes (threat cues) that indicated whether and when a brief unpleasant stimulus (sounds or pictures) might be delivered. In one experimental condition, the aversive stimuli were predictable in that they were always delivered 7 seconds after the onset of a threat cue. In another condition, the aversive stimuli were unpredictable. Participants’ fear level was assessed by the magnitude of eye-blink startle to puffs of air to the forehead. Results were that panic disorder patients and comparison subjects showed equal increases of startle from the experimental condition in which aversive sounds would never be given to the condition in which aversive sounds were given predictably. However, in the condition in which aversive sounds were given unpredictably, the increases were greater among panic disorder patients.
If one accepts unpredictable aversive sounds and pictures as analogous to unpredictable panic attacks, these results are consistent with a theoretical model of the development of panic disorder, whereby certain people progress from occasional anxiety episodes to full-fledged panic disorder because they begin to fearfully anticipate their next attack. This fearful anticipation not only prolongs anxiety between attacks but also increases the probability of attacks themselves through a positive feedback loop or vicious circle. The loop consists of fear and its sensations, which are learned cues predicting—albeit uncertainly—an imminent attack, leading to ever more fear and salient sensations. As with any research finding, more questions are raised by this article than are answered. Is vulnerability to fearing unpredictable aversive stimuli a personality risk factor for panic disorder, or is it acquired through the experience of having a series of unpredictable surges in anxiety? What would the outcome of the experiment have been if instead of aversive sounds, more panicogenic stimuli, such as inhalation of CO2-enriched air, had been delivered? Is sensitivity to unpredictable aversive stimuli specific to panic disorder, or does it also apply to generalized anxiety disorder and specific phobias?
Saying that the experiment has found a “psychophysiological marker” implies that we are dealing with a phenomenon that is more than the patient’s self-report. Although the hope of DSM anxiety disorder diagnoses has been to go beyond the patient’s words to reach a classification or an assessment of the severity of the pathology based on biology, we still, in fact, diagnose anxiety disorder patients solely on the basis of what they tell us. If physiology reflected the type and severity of pathological anxiety, we could use it as a kind of lie detector to weed out malingerers or to detect individuals who are alexithymic and deny that they are anxious when they really are. However, Grillon et al. do not claim to have such a marker, nor would their data support such a claim. Of course, if the DSM classification of anxiety proves not to have a biological basis, the search for a biological marker of DSM diagnosis will be futile. However, what Grillon et al. have accomplished is to go some distance beyond self-reporting by using startle as a measure of fear, one with somewhat different properties from the classical autonomic measures used by psychophysiologists, such as heart rate and skin conductance (2). Startle could be applied to monitor the severity of anxiety and its change after either psychological or pharmacological interventions (3). Thus, the study conducted by Grillon et al. is “translational” in that parallel experiments can be carried out in humans and nontalking animals. In animals, the fear circuits and neurochemistry of startle have been extensively charted, giving a depth of biological understanding that cannot be reached with human experimentation. Evidence has been found for a distinct neural system in “phasic fear” toward predictable shocks and in “sustained anxiety” toward unpredictable shocks (4). The former is based on the amygdala and the latter on the bed nucleus of the stria terminalis.
Do the design and results of this study give us something that would be useful in developing new treatments? The greatest hope of translation is to be able to apply findings at the molecular, physiological, or learning level—common to animals and humans—to create better treatments for pathological human anxiety. However, recently developed effective therapies for panic disorder have emphasized changing patients’ false thinking, one thing that cannot be translated between humans and animals (5). Although it is as hard to prove the validity of cognitive anxiety theories as it is to prove psychoanalytic theory (6), there is no doubt that therapies inspired by cognitive premises work. To integrate psychological and biological thinking has never been easy. Insofar as specific thoughts are the cause of panic disorder, pharmacological and anatomic localization of these thoughts are as unlikely to be useful as it would be to distinguish whether a person was thinking in French or in English. This was the point made by Szasz about conversion reactions in his controversial critique of the medical model (7). Learning paradigms that are analogous in animals and humans have been said to inspire new therapies—for example, Wolpe’s experiments in cats as a basis for systematic desensitization in human phobias (8)—but it is debatable whether animal learning experiments were scientifically necessary for developing current “behavioral” (i.e., noncognitive) interventions in anxiety disorders.
In the case of pharmacotherapies, which also benefit panic disorder but in ways other than affecting learning and thinking, translatable models take center stage. To develop new drugs for panic disorder, knowledge of what neurons, with what chemical properties and in what locations, should be altered in what way is an important guide in the design and selection of the proper molecules. This is greatly helped by the availability of a feature of human panic that is translatable to the animal level, and that is something that Grillon et al. provide in their study. Knowing that patients are hypersensitive to unpredictable but not predictable threat provides clues for potential neural targets for drug development. For example, corticotropin-releasing hormone is believed to mediate bed nucleus of the stria terminalis-mediated sustained anxiety toward unpredictable aversive events. Compounds antagonizing corticotropin-releasing hormone might someday be used to treat panic disorder (9). Full-scale clinical trials are expensive and time-consuming. Screening new compounds using fear-potentiated startle in rodents and then applying the same paradigm to humans should help accelerate drug discovery.