The Book of Woe: The DSM and the Unmaking of Psychiatry

by Gary Greenberg

  But even if the names weren’t a dead giveaway, it would be impossible to conceal the fact that personality disorders are a throwback to the Freudian conviction that our suffering reveals the neurotic shape into which our timber is twisted by the psychosexual catastrophes of childhood and the impossible demands of society, and that treatment always involves a reckoning with those forces and the person they have made us.

  The presence of personality disorders in the DSM preserves this old, officially out-of-fashion idea that personality underlies our difficulties. The book never comes out and says so, but Allen Frances wrote the personality disorders section of the DSM, and he’ll tell you unabashedly that this is the case. “I think this is the best way10 to see and treat most disorders,” he said. (Although it’s not the best way to get paid for treating them. Insurance companies don’t generally reimburse for those diagnoses, so even if you think that’s what you’re treating, you can’t tell them that.) But Frances was trained in those olden days, before the demands of science required psychiatry to distance itself from Freud by divorcing its account of our afflictions from their causes. It is as if the psychiatrists of his generation, the ones who saved their profession by meeting those demands, could not quite let go of the idea that who we are matters, and kept that thought alive on Axis II.

  • • •

  As Freud himself might have predicted, smuggling psychoanalysis into a DSM that claimed authority by purging it has created an internal conflict that yields no end of trouble.

  “I have never really been a doctor11 in the proper sense,” Freud wrote. That may be why in all his talk of neuroses and psychoses, complexes and symptoms, defenses and resistances, he remained uninterested in carving the psyche at its joints by naming its afflictions. He mostly ignored Kraepelin (who was a contemporary of his) and his parsing of mental disorders from one another. (Although he was pleased to pick up one of Kraepelin’s patients, Sergei Pankejeff, and turn him into the Wolf Man, title character of one of his most gripping case studies.) We all have parents, we all live stretched between instinct and conscience, between memory and desire, between brutality and civilization, so we all suffer degrees of the same illness; to have a personality is always to at least flirt with neurosis.

  So in the Freudian view, all personalities are more or less disordered. Which is exactly what led to those 85 percent prevalence rates, to the reliability crises, and ultimately to the suspicion that psychiatry may not belong in medicine. As problematic as the DSM’s response has become when it comes to Axis I disorders, with its categories and criteria and axes, its at least three months and its three out of six symptoms, it still works, as long as you limit your definition of works to achieves reliability under ideal conditions. But personality disorders have never worked even in that limited sense. “They have the lowest reliability12 of any major category in the book,” Frances, who served on the DSM-III personality disorders work group, told me. (The kappas were .56 to .6513.) And he’s been saying so for a long time. Within a few months of the publication of DSM-III in 1980, he was telling readers of the American Journal of Psychiatry that “the personality disorders are not at all clearly distinct14 from normal functioning or from each other,” which is why they were significantly harder for clinicians to distinguish reliably than Axis I disorders were.

  In that paper, and in one he wrote two years later, Frances suggested an alternative approach to personality disorders. “Rather than being diagnosed15 within one or another distinct personality type,” he wrote, “the patient might be rated (perhaps on a scale of 1 to 10) for each personality characteristic.” People don’t have BPD so much as they are anxious about abandonment, impulsive, and entitled—qualities they share more with people who might qualify for, say, NPD than with people whose personalities are less troublesome. Long before Darrel Regier proposed dimensional diagnosis for DSM-5, Frances was suggesting exactly that, at least for the personality disorders.

  When he started work on DSM-IV, Frances said, a dimensionalized personality disorders section was one of his goals. He said that he recruited Thomas Widiger as the DSM-IV research coordinator in part because he was a “committed dimensionalist16” and that he urged the work group to come up with an alternative to the categorical approach. “We worked hard to forge a consensus that could inform a simple DSM-IV proposal,” he said. “But we failed. We couldn’t reach agreement on how to rate the factors.” The process devolved into those pointless arguments he’d wanted to avoid—“the distinctions without differences that bedevil the field.”

  Widiger remembers it differently. He’s not so sure Frances hired him for his dimensionalist expertise—“I believe I was chosen17 to be the research coordinator because I was hardworking, conscientious, familiar with meta-analyses, and had collaborated with Allen many years before DSM-IV,” he told me, adding that only two of the one hundred projects he worked on concerned dimensions. And, he says, the problem was not really the result of contention among the troops but a failure of the general to bring them together. “There were advocates of different models,” but the same was true when DSM-III was written, “and that didn’t stop Spitzer from coming up with a compromise among them. If you recognize that the field needs to shift to a dimensional model, you can easily address the differences of opinion.”

  Twenty years later, Widiger was still mystified by Frances’s waffling, especially in light of his long history of advocacy for a dimensional approach. But Frances was committed to making changes only in the presence of incontrovertible evidence. By that principle, if the work group couldn’t resolve its own arguments, the only conclusion was that dimensions weren’t ready for the DSM, at least not his DSM. No matter how wrongheaded he thought the existing model was, no matter how long he’d been advocating for a different one, in the end he had to settle for the tepid observation, made in a section tacked on to the introduction to the personality disorders, that dimensional models were “under active investigation18,” and for a paper, published just before DSM-IV19 came out, suggesting that it was only a matter of time before personality disorders were diagnosed dimensionally. The nosological conservative had been hoist with his own petard.

  • • •

  Widiger thought he was going to get another chance. He was the sole personality disorders expert in the research planning conferences organized by Michael First in 2000. First was also an advocate of dimensions. “Patients have just one personality,” he wrote in a paper that set the DSM-5 research agenda for personality disorders. “It might be more consistent20 . . . to indicate that a patient has one personality disorder, characterized by the presence of a variety of maladaptive personality traits.” He went on to suggest a seven-point plan for implementing a dimensional approach in the DSM-5.

  First pointed out that much of the work had already been done. Just as Spitzer had drawn on years of research into diagnostic criteria in fashioning his first proposals, so too would-be DSM reformers could turn to any number of dimensional models of personality as a starting point. There was Cloninger’s Tridimensional Personality Questionnaire and Timothy Leary’s (yes, that Timothy Leary) Interpersonal Circumplex, the Big Five and the 16-PF, the DAPP and the SNAP and the SWAP. They differed in many ways, but they were all attempts to find the basic building blocks of personality, and then to show how individuals emerge as different agglomerations of those factors. The 16-PF, for instance, identifies sixteen source traits that are derived from 42 clusters, which were refined from 171 groups, which in turn were reduced from the four thousand or so adjectives in Webster’s that describe facets of personality, and the Big Five, also known as the Five Factor Model, claims that all those adjectives can be placed under one of five domains—openness to experience, conscientiousness, extraversion, agreeableness, and neuroticism. An individual personality, according to this theory, is the unique mix of these common qualities.

  No matter how they are counted and named, these factors are, of course, a
ll reifications. There really is no such thing as extraversion, even if you know it when you see it, any more than there is such a thing as Major Depressive Disorder. But these personality models have all been used extensively. Researchers have developed tests and subtests and rating scales that can locate people along the dimensions of each model. They’ve tied the resulting profiles to religious belief and political affiliation, to drug use and child abuse, to learning styles and patterns of memory. And, of course, they’ve applied their models to the personality-disordered, showing how those patients’ troubles can be reliably attributed to the lack or excess of particular factors.

  While the competing personality theories have generated as much intramural conflict as you might expect, First saw the proliferation of models as an opportunity for nosology. All that active investigation had yielded a huge body of concepts and measures from which a revision could be fashioned. And if the DSM could incorporate this data into its diagnostic regime, he thought, it could return something vital to the personality researchers who generated it: a “uniform classification of general personality functioning21” that would bring the same kind of order to the field that the DSM-III had brought to the general classification of mental illness.

  As much as everyone had to gain from this outcome, First didn’t think it would be easy to bring about. “That is why I pushed22 to have the personality research conference take place first,” he told me. “I thought if they got started working on it within the next year, before work started on the rest of DSM-5, then maybe they would have it worked out in time.”

  That meeting took place late in 2004. Tom Widiger, who had been part of the 1999 planning conferences, chaired it. Regier, he said, had asked him23 to find a “common ground” among competing personality theories and to fashion a model for DSM out of them, and he leaped at the chance to finish the work he felt Frances had abandoned. He searched for the commonalities among them, and eventually determined that the eighteen leading theories, and all their different schematics of personality, converged on four domains, each of which had its own spectrum. According to Widiger’s distillation24, we are all more or less extraverted or introverted, constrained or impulsive, emotionally stable or unstable, and antagonistic or compliant. People whose personalities were disordered could be thought of as, for instance, too extraverted or too compliant or not stable or impulsive enough, and thanks to the plethora of tests available, their location in those domains could be specified. Charlotte and Joe could thus be diagnosed without recourse to woolly concepts like borderline or narcissism; instead they could be described in terms of those factors without pretending that they had separate illnesses.

  While he knew this compromise was not a shoo-in for DSM-5, Widiger was pretty sure that he had given the APA a way to catch up with the “basic science research25 on general personality structure” and that, given the hunger for a new approach, it had a good chance for “eventual adoption.” The paradigm shift, at least when it came to personality disorders, was finally at hand. It wouldn’t be a move to a nosology based on biomarkers (although there are intriguing connections between genetics and neurochemistry on the one hand and personality traits on the other). It wouldn’t entirely solve the problem of reification, but it would at least rid one section of the DSM of diagnostic categories that everyone agreed really didn’t make sense and replace them with dimensions, which nearly everyone agreed were a better way to conceptualize mental illness. The personality disorders work group—with Widiger at its helm, or so he and many others thought—would lead psychiatry into the future.

  • • •

  “The devil, of course26, could be in the details,” Widiger wrote at the end of his “common ground” paper. The details he had in mind were scientific, and, given his prominent position in the field, he figured he would be heading the team that would work them out. But the details that arose were not scientific at all. In 2005, he and a colleague began to plan another conference about dimensions in DSM-5. Three months into those conversations, Widiger discovered that he would not be invited to that conference. And then, late in 2006, psychiatrist John Livesley, who had attended the conference, called to tell Widiger that he had just seen the list of people who would be on the DSM-5 work group. Widiger had gotten the Michael First treatment—an acknowledged leader in the field, who had done yeoman service for the APA, who most people thought would get the job, had been unceremoniously exiled. “Nobody on the work group27 ever asked for my input or informed me about what was going on,” Widiger said. He was left to figure it out from what was posted on the website.

  The first proposals Widiger saw appeared in February 2010. They represented, according to the work group, a “significant reformulation.” Diagnosis would no longer be a matter of checking the criteria against the patient. But neither would it be a shift to a purely dimensional approach. Instead, they proposed a “hybrid” model that required clinicians to go through a four-step procedure. First, they would use seven criteria, including identity integration and cooperativeness, to determine whether the patient had a personality disorder, which was defined as the failure to develop “a sense of self-identity28 and the capacity for interpersonal function.” They would then determine which of five types of personality disorder (antisocial/psychopathic, avoidant, borderline, obsessive-compulsive, or schizotypal) the patient had. To figure this out, they would not use criteria, but instead match the patient to prototypical descriptions of the disorder. Then they would see which of thirty-seven trait facets—social withdrawal, for example, or recklessness—nested in six trait domains, such as negative emotionality and disinhibition, best described the patient. If the patient didn’t fit one of the types, they would diagnose her with a personality disorder, and the facets and domains would be used to describe the disorder further. And finally they would determine how badly disordered the patient’s personality was by rating it on a scale of one to three along four dimensions for two different areas of functioning.

  Don’t worry if you didn’t get that. Some hybrids are elegant (think of Priuses or Gala apples). Others are more like mules. Even if the personality disorders proposal makes sense (and I’m not sure it does), as a clinician I can assure you that it is way too complicated and time-consuming for anyone to actually use. More important, however, the proposal had little if anything to do with all those validated theories and their measurement instruments that Widiger had labored so hard to integrate into a single model. The trait facets and domains seemed to bear little relationship to the long history of research that had once made personality disorders seem so promising as a game-changer. And the types—well, it was at least clear where the committee had found them: in the DSM-IV. But they didn’t explain why they had adopted only five of the original ten personality disorders.

  Actually, they did try to explain. Andrew Skodol, the work group chair, reported that they had conducted a literature review and eliminated the five diagnoses that the least amount of research had been done on. They argued that the lack of studies indicated that these diagnoses had insufficient “empirical evidence29 of validity and clinical utility,” failing to note that absence of evidence was not evidence of absence, that, in fact, the decision amounted to letting the marketplace decide which diagnoses should stay and which should go.

  Skodol also ventured an explanation for not simply choosing one of the established personality theories, with its validated concepts and measures, and instead cobbling together a model that, as the APA work group noted on the website, was still “pending empirical validation.” “We knew we couldn’t incorporate30 anybody’s model in toto,” he told me, “because the APA wasn’t going to pay royalties” to any copyright holder.

  Whatever the rationale, the proposal left Widiger hopping mad. “The empirical support is a farce,” he wrote in an e-mail to his colleagues, the section “embarrassingly bad31.” It would become “the laughingstock of the DSM and psychiatry,” he predicted. “It’s a monumenta
l mistake.”

  Widiger wasn’t alone in being appalled. John Livesley went public with his charge that the proposal was “a disaster.” He withdrew his name from the work group’s journal article describing the changes and published his own dissenting opinion in March 2010, lamenting the proposal as a “lost opportunity” that “negates years of progress32 toward a scientific-based classification system.” In April, Frances called the proposal a “cumbersome hodgepodge33” whose dimensions were “the favorites of the few members of the work group who have personal research experience in this area.” The comments submitted to the website during the spring ran strongly against the proposals. In September, a group of eight psychiatrists and psychologists, some of the biggest names in the business, ranging from the ancient psychoanalyst Otto Kernberg to the even more ancient father of cognitive-behavioral therapy, Aaron Beck, to the upstart political psychologist Drew Westen, put aside their differences to tell readers of the American Journal of Psychiatry that the proposal was “an unwieldy conglomeration34” whose traits and facets “have not been investigated empirically,” which meant that the model “no longer rests on the decades of research” that had been “the chief rationale” for making the change to a dimensional model in the first place.

  In the fall of 2010, the task force staged what Skodol called an “intervention.” “They hated the prototypes”—the descriptions of personality disorders that had replaced the criteria. “They reminded them of DSM-II,” he explained, and they weren’t crazy about the rest of it, either. “You guys look like outliers,” they told him, as they ordered the group back to the drawing board—this time with two “consultants” (“minders,” one committee member called them) to assist in reformulating the proposals. In a series of weekly phone conferences, chaired not by Skodol but by one of the consultants, a new proposal was hammered out and a “simplified and streamlined” proposal was posted on the website in late June. Now there were five trait domains instead of six, and thirty-one facets instead of thirty-seven, and six types instead of five. The criteria for those types were back, reflecting the task force’s mandate to, as Skodol put it, “make the section look more like the rest of them.”