Redefining personality disorders: Proposed revisions for DSM-5

A major update to the diagnostic manual used by mental health clinicians around the world is expected to inspire lively debate. Proposed revisions to the personality disorders (PD) section of the next edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), which is scheduled to be published in 2013, have generated great controversy because they would introduce a dimensional model to the categorical system and 4 PDs would be eliminated.

“The importance of personality functioning and personality traits is the major innovation here,” said Andrew Skodol, MD, the DSM-5 Personality and Personality Disorders Work Group’s chair and a Research Professor of Psychiatry at the University of Arizona College of Medicine. “In the past, we viewed personality disorders as binary. You either had one or you didn’t. But we now understand that personality pathology is a matter of degree.”¹

Mark Zimmerman, MD, has written several papers—some of which are in press—about how these revisions might impact clinicians and whether the revisions are necessary. He is Director of the Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project, an ongoing clinical research study involving the integration of research assessment methods into clinical practice.

Proposed revisions, rationale, and literature reviews for DSM-5 are available at www.DSM5.org and anyone, including the general public, was invited to provide feedback through the Web site. Current Psychiatry Associate Editor Donald W. Black, MD, interviewed Dr. Zimmerman on June 29, 2011, just a few days after the latest proposed revision was posted on June 21, 2011.

DR. BLACK: What is your understanding of the DSM-5 Personality Disorders Work Group proposal to revamp the PD category?

DR. ZIMMERMAN: The initial proposal, released in February 2010, was complex and generated a fair amount of critical commentary related to the marked changes in the approach toward diagnosis of PDs. That proposal replaced diagnostic criteria with a prototype description of personality types that patients would need to match. It also eliminated 5 PDs—paranoid, schizoid, histrionic, dependent, and narcissistic—retained antisocial, avoidant, borderline, obsessive-compulsive, and schizotypal, and introduced trait level ratings. The June 21 revision proposes eliminating only 4 disorders—narcissistic was retained—and the Work Group is no longer suggesting using prototypes but instead have diagnostic criteria (Table).^2,3 We do not know if this is the final proposal because similar to the first proposal, it is not presented with much supporting empirical evidence that demonstrates its superiority toward diagnosing PDs compared with the DSM-IV approach.

Table

Personality disorder criteria: DSM-IV vs DSM-5

I’m not suggesting that the DSM-IV approach is without problems. My attitude is that before going forward with a change to the official diagnostic nomenclature, you need to clearly establish that the new way of doing things is better than the previous way by whatever metric you use.

DR. BLACK: Do you believe there is a need to revamp or revise the DSM-IV PD criteria?

DR. ZIMMERMAN: I think a number of the arguments put forth by the DSM-5 Work Group as justifications for revising the criteria do not hold up to empirical study.

One of the issues is the argument that there’s too much comorbidity among PDs. The theory is that disorders are not unique diagnostic entities if they are so frequently comorbid with other disorders. But how much comorbidity is too much? The DSM-5 Work Group doesn’t say. Oldham et al⁴ found comorbidity rates of 70% to 90%, depending on which semi-structured diagnostic interview was used; however, this was among individuals presenting for psychodynamic treatment of PDs.

I wanted to look at the comorbidity rates in nontreatment-seeking samples to find out if treatment seeking is associated with comorbidity. I reviewed the literature and identified 7 general population epidemiological studies that presented data on the number of individuals with ≥2 PD diagnoses. In these studies, the comorbidity rate is approximately 25%, which is one-half or less than the rates found in patient populations.⁵ This finding suggested to me that this may not be a nosology problem unless you think 25% comorbidity is too high. The DSM-5 people don’t speak to that, although quite frankly with 10 PDs I don’t think the 25% comorbidity rate is excessive. However, a comorbidity rate of 25% was much lower than that found in patient samples and suggests to me that one of the primary stated reasons of deleting 4 PDs may not be valid.

DR. BLACK: Assuming there is a need to revise the PD section, how would you have gone about that process?

DR. ZIMMERMAN: Whatever deficiencies you perceive in the criteria, the process should be that you come up with an alternative, examine the alternative empirically, and this is followed by independent replication that the new approach is superior to the prior one. My view is that it is not sufficient justification to make a change because there is a problem with the prior approach.

We can argue as to whether there really are problems with, for example, the categorical nature of classification. My research group and I wrote a paper arguing that DSM-IV can be interpreted as having a dimensional component (Box).⁶ DSM-IV suggests that clinicians record on axis II that a patient has some traits of a disorder even when the full criteria are not met. With that in mind, we conceptualized DSM-IV as having a 3-point dimension, where 0 means no traits of the disorder, 1 indicates subthreshold traits, and 2 indicates that the disorder is present. In a study of >2,000 patients, we found that DSM-IV’s 3-point dimensional approach was as highly associated with measures of psychosocial morbidity as more finely graded dimensional systems.⁶ We therefore concluded that DSM-IV already includes a dimensional system and questioned why we need to change that approach.

Box

One of my concerns with the dimensional system as currently proposed is the uncertain significance and possible implications of someone being given a low, non-zero rating. How might this play out in a custody evaluation of someone who is said to be “a little borderline”? What might the implications of non-zero ratings be in obtaining life insurance? The potential practical consequences of low ratings have not, to my knowledge, been discussed. Because of this concern we decided to do a study to determine if there was any clinical significance to low dimensional scores. I had hypothesized that if we compared individuals who had no criteria and only 1 BPD criterion, there would be no difference.

To be frank, I was seeking to show that there was no validity to low levels of pathology and therefore the DSM-5 group probably is getting into dangerous territory. In fact, we found that there were rather significant and robust differences between individuals with 0 criteria and 1 criterion.⁷ Even though this finding didn’t support my hypothesis, I thought it was significant because it supported the DSM-5 Work Group and I felt compelled to publish that data.

We now had 2 interesting pieces of information. A few years ago we published a study on borderline personality disorder (BPD) that found once you hit the diagnostic threshold it made no difference how many criteria you met.⁸ On the other hand, when you were below the diagnostic threshold, having 1 criterion vs 0 made a big difference. In addition, a fair number of studies show that dimensional models capture more of the variance in personality pathology than categorical variables.^9-12

This lead to our next study in which we hypothesized that dimensionality was only important when the person didn’t meet criteria, not when they did meet criteria.¹³ So we divided patients in the MIDAS study into those with 0 to 4 BPD criteria and those with ≥5 and counted the number of criteria that were met. Then we correlated each of those 2 dimensional scores with various indicators of illness severity, such as number of suicide attempts, number of psychiatric hospitalizations, and amount of time missed from work in the past 5 years. We found that for individuals who already achieved the diagnostic threshold there were very low correlations with these psychosocial morbidity variables. But for patients with subthreshold symptomatology, there were significant correlations and those correlations were significantly higher than the correlations for the other group. We therefore suggested that dimensionality is important but only when you don’t meet the diagnostic threshold. Thus, we came to the conclusion that DSM-IV already provides for capturing the important dimensional nature of PDs.

DR. BLACK: I’ve discussed this issue with a number of people who basically say doctors tend to think categorically, they don’t think along dimensions. Would it be difficult for psychiatrists to accept this type of system because it’s so different from how physicians are trained to think?

DR. ZIMMERMAN: I think doctors do think categorically and about traits, not necessarily disorders. For example, we’ll see a patient and a clinician will say he’s overly perfectionistic, but there’s no perfectionistic disorder in DSM-IV. This patient may or may not have obsessive-compulsive personality disorder.

I think assessment and diagnosis in routine clinical practice are not nearly as comprehensive as in research. I think psychiatrists often are picking up on traits that they think are clinically significant, but even within that context, they’re thinking categorically, that the patient is perfectionistic rather than rating him a 7 on a scale from 0 to 10 in terms of perfectionism.

Eliminated disorders

DR. BLACK: The proposal will cut the number of PDs to 6 plus personality disorder trait specified and those remaining are to be called types. How did the DSM-5 Work Group select the 5 (now 4) disorders to get rid of? Did they just pick ones that were infrequently used?

DR. ZIMMERMAN: They retained the disorders that were studied in the Collaborative Longitudinal Personality Disorders Study study¹⁴ plus others with well established validity.

DR. BLACK: What do you think about that plan to reduce the number of PDs?

DR. ZIMMERMAN: The biggest problem I have is that the DSM-5 Work Group didn’t present any data on the implications of their plan. The conceptual justification was to reduce comorbidity rates. Well, you can hypothesize that comorbidity would actually increase if you retained only those disorders that are more frequently comorbid with other disorders. Would there be any individuals who only have 1 of the excluded diagnoses? Would there be false negatives? They didn’t indicate whether comorbidity would drop and by how much. And they didn’t indicate if there would be a potential impact on missing cases.

We did such an analysis because we had the data set available from the MIDAS project.¹⁵ We wanted to know if you excluded the 5 diagnoses that (at the time) were proposed for exclusion—narcissistic, paranoid, schizoid, dependent, and histrionic—what percentage of individuals would no longer be diagnosed with a PD? Second, how much would comorbidity rates change? And third, how did individuals who would no longer be diagnosed with a PD compare with individuals who never had a PD?

We found that the comorbidity rates did, in fact, drop from 30% to 21%. We found that the rate of PDs dropped only a little, but approximately 10% of individuals who previously would have been diagnosed with a PD would no longer be diagnosed. We compared individuals in the excluded group—those who had only 1 of the PDs that would no longer be considered a PD—with a group of patients who had a retained PD and also compared them to individuals with no PD. We found that the retained PD group and the excluded group did not differ on measures of psychosocial morbidity, such as Global Assessment of Functioning scores, hospitalizations, suicidality, number of current axis I disorders, etc. Also, the excluded group clearly was different than the no PD group. We questioned whether or not those in the excluded group might end up being false negative diagnoses in DSM-5. Certainly DSM-5 provides a provision to use trait ratings to still diagnose a PD, called personality disorder trait specified, which would be somewhat analogous to PD not otherwise specified (NOS).

It’s ironic insofar as another of the issues considered by the DSM-5 Work Group to be a problem with axis II is lack of coverage and that too many individuals are diagnosed with PD NOS. Their proposal to exclude PDs could result in more individuals being diagnosed with PD NOS. I know the group would disagree with that perspective, but they provided no evidence to support its view.

As I said at the beginning of this interview, I think we should be talking about this from a scientific perspective and nothing more than that.

Related Resource

• Widiger TA, Simonsen E, Sirovatka PJ, et al. Dimensional models of personality disorders: Refining the research agenda for DSM-V. Arlington, VA: American Psychiatric Publishing, Inc; 2007.

Disclosures

• Dr. Zimmerman reports no financial relationship with any company whose products are mentioned in this article or with manufacturers of competing products.
• Dr. Black receives research support from AstraZeneca and Psyadon.

A major update to the diagnostic manual used by mental health clinicians around the world is expected to inspire lively debate. Proposed revisions to the personality disorders (PD) section of the next edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), which is scheduled to be published in 2013, have generated great controversy because they would introduce a dimensional model to the categorical system and 4 PDs would be eliminated.

“The importance of personality functioning and personality traits is the major innovation here,” said Andrew Skodol, MD, the DSM-5 Personality and Personality Disorders Work Group’s chair and a Research Professor of Psychiatry at the University of Arizona College of Medicine. “In the past, we viewed personality disorders as binary. You either had one or you didn’t. But we now understand that personality pathology is a matter of degree.”¹

Mark Zimmerman, MD, has written several papers—some of which are in press—about how these revisions might impact clinicians and whether the revisions are necessary. He is Director of the Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project, an ongoing clinical research study involving the integration of research assessment methods into clinical practice.

Proposed revisions, rationale, and literature reviews for DSM-5 are available at www.DSM5.org and anyone, including the general public, was invited to provide feedback through the Web site. Current Psychiatry Associate Editor Donald W. Black, MD, interviewed Dr. Zimmerman on June 29, 2011, just a few days after the latest proposed revision was posted on June 21, 2011.

DR. BLACK: What is your understanding of the DSM-5 Personality Disorders Work Group proposal to revamp the PD category?

DR. ZIMMERMAN: The initial proposal, released in February 2010, was complex and generated a fair amount of critical commentary related to the marked changes in the approach toward diagnosis of PDs. That proposal replaced diagnostic criteria with a prototype description of personality types that patients would need to match. It also eliminated 5 PDs—paranoid, schizoid, histrionic, dependent, and narcissistic—retained antisocial, avoidant, borderline, obsessive-compulsive, and schizotypal, and introduced trait level ratings. The June 21 revision proposes eliminating only 4 disorders—narcissistic was retained—and the Work Group is no longer suggesting using prototypes but instead have diagnostic criteria (Table).^2,3 We do not know if this is the final proposal because similar to the first proposal, it is not presented with much supporting empirical evidence that demonstrates its superiority toward diagnosing PDs compared with the DSM-IV approach.

Table

Personality disorder criteria: DSM-IV vs DSM-5

I’m not suggesting that the DSM-IV approach is without problems. My attitude is that before going forward with a change to the official diagnostic nomenclature, you need to clearly establish that the new way of doing things is better than the previous way by whatever metric you use.

DR. BLACK: Do you believe there is a need to revamp or revise the DSM-IV PD criteria?

DR. ZIMMERMAN: I think a number of the arguments put forth by the DSM-5 Work Group as justifications for revising the criteria do not hold up to empirical study.

One of the issues is the argument that there’s too much comorbidity among PDs. The theory is that disorders are not unique diagnostic entities if they are so frequently comorbid with other disorders. But how much comorbidity is too much? The DSM-5 Work Group doesn’t say. Oldham et al⁴ found comorbidity rates of 70% to 90%, depending on which semi-structured diagnostic interview was used; however, this was among individuals presenting for psychodynamic treatment of PDs.

I wanted to look at the comorbidity rates in nontreatment-seeking samples to find out if treatment seeking is associated with comorbidity. I reviewed the literature and identified 7 general population epidemiological studies that presented data on the number of individuals with ≥2 PD diagnoses. In these studies, the comorbidity rate is approximately 25%, which is one-half or less than the rates found in patient populations.⁵ This finding suggested to me that this may not be a nosology problem unless you think 25% comorbidity is too high. The DSM-5 people don’t speak to that, although quite frankly with 10 PDs I don’t think the 25% comorbidity rate is excessive. However, a comorbidity rate of 25% was much lower than that found in patient samples and suggests to me that one of the primary stated reasons of deleting 4 PDs may not be valid.

DR. BLACK: Assuming there is a need to revise the PD section, how would you have gone about that process?

DR. ZIMMERMAN: Whatever deficiencies you perceive in the criteria, the process should be that you come up with an alternative, examine the alternative empirically, and this is followed by independent replication that the new approach is superior to the prior one. My view is that it is not sufficient justification to make a change because there is a problem with the prior approach.

We can argue as to whether there really are problems with, for example, the categorical nature of classification. My research group and I wrote a paper arguing that DSM-IV can be interpreted as having a dimensional component (Box).⁶ DSM-IV suggests that clinicians record on axis II that a patient has some traits of a disorder even when the full criteria are not met. With that in mind, we conceptualized DSM-IV as having a 3-point dimension, where 0 means no traits of the disorder, 1 indicates subthreshold traits, and 2 indicates that the disorder is present. In a study of >2,000 patients, we found that DSM-IV’s 3-point dimensional approach was as highly associated with measures of psychosocial morbidity as more finely graded dimensional systems.⁶ We therefore concluded that DSM-IV already includes a dimensional system and questioned why we need to change that approach.

Box

One of my concerns with the dimensional system as currently proposed is the uncertain significance and possible implications of someone being given a low, non-zero rating. How might this play out in a custody evaluation of someone who is said to be “a little borderline”? What might the implications of non-zero ratings be in obtaining life insurance? The potential practical consequences of low ratings have not, to my knowledge, been discussed. Because of this concern we decided to do a study to determine if there was any clinical significance to low dimensional scores. I had hypothesized that if we compared individuals who had no criteria and only 1 BPD criterion, there would be no difference.

To be frank, I was seeking to show that there was no validity to low levels of pathology and therefore the DSM-5 group probably is getting into dangerous territory. In fact, we found that there were rather significant and robust differences between individuals with 0 criteria and 1 criterion.⁷ Even though this finding didn’t support my hypothesis, I thought it was significant because it supported the DSM-5 Work Group and I felt compelled to publish that data.

We now had 2 interesting pieces of information. A few years ago we published a study on borderline personality disorder (BPD) that found once you hit the diagnostic threshold it made no difference how many criteria you met.⁸ On the other hand, when you were below the diagnostic threshold, having 1 criterion vs 0 made a big difference. In addition, a fair number of studies show that dimensional models capture more of the variance in personality pathology than categorical variables.^9-12

This lead to our next study in which we hypothesized that dimensionality was only important when the person didn’t meet criteria, not when they did meet criteria.¹³ So we divided patients in the MIDAS study into those with 0 to 4 BPD criteria and those with ≥5 and counted the number of criteria that were met. Then we correlated each of those 2 dimensional scores with various indicators of illness severity, such as number of suicide attempts, number of psychiatric hospitalizations, and amount of time missed from work in the past 5 years. We found that for individuals who already achieved the diagnostic threshold there were very low correlations with these psychosocial morbidity variables. But for patients with subthreshold symptomatology, there were significant correlations and those correlations were significantly higher than the correlations for the other group. We therefore suggested that dimensionality is important but only when you don’t meet the diagnostic threshold. Thus, we came to the conclusion that DSM-IV already provides for capturing the important dimensional nature of PDs.

DR. BLACK: I’ve discussed this issue with a number of people who basically say doctors tend to think categorically, they don’t think along dimensions. Would it be difficult for psychiatrists to accept this type of system because it’s so different from how physicians are trained to think?

DR. ZIMMERMAN: I think doctors do think categorically and about traits, not necessarily disorders. For example, we’ll see a patient and a clinician will say he’s overly perfectionistic, but there’s no perfectionistic disorder in DSM-IV. This patient may or may not have obsessive-compulsive personality disorder.

I think assessment and diagnosis in routine clinical practice are not nearly as comprehensive as in research. I think psychiatrists often are picking up on traits that they think are clinically significant, but even within that context, they’re thinking categorically, that the patient is perfectionistic rather than rating him a 7 on a scale from 0 to 10 in terms of perfectionism.

Eliminated disorders

DR. BLACK: The proposal will cut the number of PDs to 6 plus personality disorder trait specified and those remaining are to be called types. How did the DSM-5 Work Group select the 5 (now 4) disorders to get rid of? Did they just pick ones that were infrequently used?

DR. ZIMMERMAN: They retained the disorders that were studied in the Collaborative Longitudinal Personality Disorders Study study¹⁴ plus others with well established validity.

DR. BLACK: What do you think about that plan to reduce the number of PDs?

DR. ZIMMERMAN: The biggest problem I have is that the DSM-5 Work Group didn’t present any data on the implications of their plan. The conceptual justification was to reduce comorbidity rates. Well, you can hypothesize that comorbidity would actually increase if you retained only those disorders that are more frequently comorbid with other disorders. Would there be any individuals who only have 1 of the excluded diagnoses? Would there be false negatives? They didn’t indicate whether comorbidity would drop and by how much. And they didn’t indicate if there would be a potential impact on missing cases.

We did such an analysis because we had the data set available from the MIDAS project.¹⁵ We wanted to know if you excluded the 5 diagnoses that (at the time) were proposed for exclusion—narcissistic, paranoid, schizoid, dependent, and histrionic—what percentage of individuals would no longer be diagnosed with a PD? Second, how much would comorbidity rates change? And third, how did individuals who would no longer be diagnosed with a PD compare with individuals who never had a PD?

We found that the comorbidity rates did, in fact, drop from 30% to 21%. We found that the rate of PDs dropped only a little, but approximately 10% of individuals who previously would have been diagnosed with a PD would no longer be diagnosed. We compared individuals in the excluded group—those who had only 1 of the PDs that would no longer be considered a PD—with a group of patients who had a retained PD and also compared them to individuals with no PD. We found that the retained PD group and the excluded group did not differ on measures of psychosocial morbidity, such as Global Assessment of Functioning scores, hospitalizations, suicidality, number of current axis I disorders, etc. Also, the excluded group clearly was different than the no PD group. We questioned whether or not those in the excluded group might end up being false negative diagnoses in DSM-5. Certainly DSM-5 provides a provision to use trait ratings to still diagnose a PD, called personality disorder trait specified, which would be somewhat analogous to PD not otherwise specified (NOS).

It’s ironic insofar as another of the issues considered by the DSM-5 Work Group to be a problem with axis II is lack of coverage and that too many individuals are diagnosed with PD NOS. Their proposal to exclude PDs could result in more individuals being diagnosed with PD NOS. I know the group would disagree with that perspective, but they provided no evidence to support its view.

As I said at the beginning of this interview, I think we should be talking about this from a scientific perspective and nothing more than that.

Related Resource

• Widiger TA, Simonsen E, Sirovatka PJ, et al. Dimensional models of personality disorders: Refining the research agenda for DSM-V. Arlington, VA: American Psychiatric Publishing, Inc; 2007.

Disclosures

• Dr. Zimmerman reports no financial relationship with any company whose products are mentioned in this article or with manufacturers of competing products.
• Dr. Black receives research support from AstraZeneca and Psyadon.

A major update to the diagnostic manual used by mental health clinicians around the world is expected to inspire lively debate. Proposed revisions to the personality disorders (PD) section of the next edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-5), which is scheduled to be published in 2013, have generated great controversy because they would introduce a dimensional model to the categorical system and 4 PDs would be eliminated.

“The importance of personality functioning and personality traits is the major innovation here,” said Andrew Skodol, MD, the DSM-5 Personality and Personality Disorders Work Group’s chair and a Research Professor of Psychiatry at the University of Arizona College of Medicine. “In the past, we viewed personality disorders as binary. You either had one or you didn’t. But we now understand that personality pathology is a matter of degree.”¹

Mark Zimmerman, MD, has written several papers—some of which are in press—about how these revisions might impact clinicians and whether the revisions are necessary. He is Director of the Rhode Island Methods to Improve Diagnostic Assessment and Services (MIDAS) project, an ongoing clinical research study involving the integration of research assessment methods into clinical practice.

Proposed revisions, rationale, and literature reviews for DSM-5 are available at www.DSM5.org and anyone, including the general public, was invited to provide feedback through the Web site. Current Psychiatry Associate Editor Donald W. Black, MD, interviewed Dr. Zimmerman on June 29, 2011, just a few days after the latest proposed revision was posted on June 21, 2011.

DR. BLACK: What is your understanding of the DSM-5 Personality Disorders Work Group proposal to revamp the PD category?

DR. ZIMMERMAN: The initial proposal, released in February 2010, was complex and generated a fair amount of critical commentary related to the marked changes in the approach toward diagnosis of PDs. That proposal replaced diagnostic criteria with a prototype description of personality types that patients would need to match. It also eliminated 5 PDs—paranoid, schizoid, histrionic, dependent, and narcissistic—retained antisocial, avoidant, borderline, obsessive-compulsive, and schizotypal, and introduced trait level ratings. The June 21 revision proposes eliminating only 4 disorders—narcissistic was retained—and the Work Group is no longer suggesting using prototypes but instead have diagnostic criteria (Table).^2,3 We do not know if this is the final proposal because similar to the first proposal, it is not presented with much supporting empirical evidence that demonstrates its superiority toward diagnosing PDs compared with the DSM-IV approach.

Table

Personality disorder criteria: DSM-IV vs DSM-5

I’m not suggesting that the DSM-IV approach is without problems. My attitude is that before going forward with a change to the official diagnostic nomenclature, you need to clearly establish that the new way of doing things is better than the previous way by whatever metric you use.

DR. BLACK: Do you believe there is a need to revamp or revise the DSM-IV PD criteria?

DR. ZIMMERMAN: I think a number of the arguments put forth by the DSM-5 Work Group as justifications for revising the criteria do not hold up to empirical study.

One of the issues is the argument that there’s too much comorbidity among PDs. The theory is that disorders are not unique diagnostic entities if they are so frequently comorbid with other disorders. But how much comorbidity is too much? The DSM-5 Work Group doesn’t say. Oldham et al⁴ found comorbidity rates of 70% to 90%, depending on which semi-structured diagnostic interview was used; however, this was among individuals presenting for psychodynamic treatment of PDs.

I wanted to look at the comorbidity rates in nontreatment-seeking samples to find out if treatment seeking is associated with comorbidity. I reviewed the literature and identified 7 general population epidemiological studies that presented data on the number of individuals with ≥2 PD diagnoses. In these studies, the comorbidity rate is approximately 25%, which is one-half or less than the rates found in patient populations.⁵ This finding suggested to me that this may not be a nosology problem unless you think 25% comorbidity is too high. The DSM-5 people don’t speak to that, although quite frankly with 10 PDs I don’t think the 25% comorbidity rate is excessive. However, a comorbidity rate of 25% was much lower than that found in patient samples and suggests to me that one of the primary stated reasons of deleting 4 PDs may not be valid.

DR. BLACK: Assuming there is a need to revise the PD section, how would you have gone about that process?

DR. ZIMMERMAN: Whatever deficiencies you perceive in the criteria, the process should be that you come up with an alternative, examine the alternative empirically, and this is followed by independent replication that the new approach is superior to the prior one. My view is that it is not sufficient justification to make a change because there is a problem with the prior approach.

We can argue as to whether there really are problems with, for example, the categorical nature of classification. My research group and I wrote a paper arguing that DSM-IV can be interpreted as having a dimensional component (Box).⁶ DSM-IV suggests that clinicians record on axis II that a patient has some traits of a disorder even when the full criteria are not met. With that in mind, we conceptualized DSM-IV as having a 3-point dimension, where 0 means no traits of the disorder, 1 indicates subthreshold traits, and 2 indicates that the disorder is present. In a study of >2,000 patients, we found that DSM-IV’s 3-point dimensional approach was as highly associated with measures of psychosocial morbidity as more finely graded dimensional systems.⁶ We therefore concluded that DSM-IV already includes a dimensional system and questioned why we need to change that approach.

Box

One of my concerns with the dimensional system as currently proposed is the uncertain significance and possible implications of someone being given a low, non-zero rating. How might this play out in a custody evaluation of someone who is said to be “a little borderline”? What might the implications of non-zero ratings be in obtaining life insurance? The potential practical consequences of low ratings have not, to my knowledge, been discussed. Because of this concern we decided to do a study to determine if there was any clinical significance to low dimensional scores. I had hypothesized that if we compared individuals who had no criteria and only 1 BPD criterion, there would be no difference.

To be frank, I was seeking to show that there was no validity to low levels of pathology and therefore the DSM-5 group probably is getting into dangerous territory. In fact, we found that there were rather significant and robust differences between individuals with 0 criteria and 1 criterion.⁷ Even though this finding didn’t support my hypothesis, I thought it was significant because it supported the DSM-5 Work Group and I felt compelled to publish that data.

We now had 2 interesting pieces of information. A few years ago we published a study on borderline personality disorder (BPD) that found once you hit the diagnostic threshold it made no difference how many criteria you met.⁸ On the other hand, when you were below the diagnostic threshold, having 1 criterion vs 0 made a big difference. In addition, a fair number of studies show that dimensional models capture more of the variance in personality pathology than categorical variables.^9-12

This lead to our next study in which we hypothesized that dimensionality was only important when the person didn’t meet criteria, not when they did meet criteria.¹³ So we divided patients in the MIDAS study into those with 0 to 4 BPD criteria and those with ≥5 and counted the number of criteria that were met. Then we correlated each of those 2 dimensional scores with various indicators of illness severity, such as number of suicide attempts, number of psychiatric hospitalizations, and amount of time missed from work in the past 5 years. We found that for individuals who already achieved the diagnostic threshold there were very low correlations with these psychosocial morbidity variables. But for patients with subthreshold symptomatology, there were significant correlations and those correlations were significantly higher than the correlations for the other group. We therefore suggested that dimensionality is important but only when you don’t meet the diagnostic threshold. Thus, we came to the conclusion that DSM-IV already provides for capturing the important dimensional nature of PDs.

DR. BLACK: I’ve discussed this issue with a number of people who basically say doctors tend to think categorically, they don’t think along dimensions. Would it be difficult for psychiatrists to accept this type of system because it’s so different from how physicians are trained to think?

DR. ZIMMERMAN: I think doctors do think categorically and about traits, not necessarily disorders. For example, we’ll see a patient and a clinician will say he’s overly perfectionistic, but there’s no perfectionistic disorder in DSM-IV. This patient may or may not have obsessive-compulsive personality disorder.

I think assessment and diagnosis in routine clinical practice are not nearly as comprehensive as in research. I think psychiatrists often are picking up on traits that they think are clinically significant, but even within that context, they’re thinking categorically, that the patient is perfectionistic rather than rating him a 7 on a scale from 0 to 10 in terms of perfectionism.

Eliminated disorders

DR. BLACK: The proposal will cut the number of PDs to 6 plus personality disorder trait specified and those remaining are to be called types. How did the DSM-5 Work Group select the 5 (now 4) disorders to get rid of? Did they just pick ones that were infrequently used?

DR. ZIMMERMAN: They retained the disorders that were studied in the Collaborative Longitudinal Personality Disorders Study study¹⁴ plus others with well established validity.

DR. BLACK: What do you think about that plan to reduce the number of PDs?

DR. ZIMMERMAN: The biggest problem I have is that the DSM-5 Work Group didn’t present any data on the implications of their plan. The conceptual justification was to reduce comorbidity rates. Well, you can hypothesize that comorbidity would actually increase if you retained only those disorders that are more frequently comorbid with other disorders. Would there be any individuals who only have 1 of the excluded diagnoses? Would there be false negatives? They didn’t indicate whether comorbidity would drop and by how much. And they didn’t indicate if there would be a potential impact on missing cases.

We did such an analysis because we had the data set available from the MIDAS project.¹⁵ We wanted to know if you excluded the 5 diagnoses that (at the time) were proposed for exclusion—narcissistic, paranoid, schizoid, dependent, and histrionic—what percentage of individuals would no longer be diagnosed with a PD? Second, how much would comorbidity rates change? And third, how did individuals who would no longer be diagnosed with a PD compare with individuals who never had a PD?

We found that the comorbidity rates did, in fact, drop from 30% to 21%. We found that the rate of PDs dropped only a little, but approximately 10% of individuals who previously would have been diagnosed with a PD would no longer be diagnosed. We compared individuals in the excluded group—those who had only 1 of the PDs that would no longer be considered a PD—with a group of patients who had a retained PD and also compared them to individuals with no PD. We found that the retained PD group and the excluded group did not differ on measures of psychosocial morbidity, such as Global Assessment of Functioning scores, hospitalizations, suicidality, number of current axis I disorders, etc. Also, the excluded group clearly was different than the no PD group. We questioned whether or not those in the excluded group might end up being false negative diagnoses in DSM-5. Certainly DSM-5 provides a provision to use trait ratings to still diagnose a PD, called personality disorder trait specified, which would be somewhat analogous to PD not otherwise specified (NOS).

It’s ironic insofar as another of the issues considered by the DSM-5 Work Group to be a problem with axis II is lack of coverage and that too many individuals are diagnosed with PD NOS. Their proposal to exclude PDs could result in more individuals being diagnosed with PD NOS. I know the group would disagree with that perspective, but they provided no evidence to support its view.

As I said at the beginning of this interview, I think we should be talking about this from a scientific perspective and nothing more than that.

Related Resource

• Widiger TA, Simonsen E, Sirovatka PJ, et al. Dimensional models of personality disorders: Refining the research agenda for DSM-V. Arlington, VA: American Psychiatric Publishing, Inc; 2007.

Disclosures

• Dr. Zimmerman reports no financial relationship with any company whose products are mentioned in this article or with manufacturers of competing products.
• Dr. Black receives research support from AstraZeneca and Psyadon.