The concept of mood temperaments is perhaps the key issue of distinction in relation to personality “disorders.” It is another common misconception among psychoanalytically oriented clinicians to think about personality only in terms of personality “disorders.” It is perhaps an irrefutable scientific fact that there are orders of magnitude more scientific data for validity of personality traits than for DSM-defined personality disorder [53]. Despite this clear scientific evidence, the DSM-5 process ended with a refusal to allow personality traits into the psychiatric diagnostic system, whereas the prior, poorly scientifically validated, structure of personality disorder was retained with minor alterations.

It is scientifically more sound to recognize, however, that one first can address personality with the concept of personality traits, among which a number are replicated with sometimes varying designations for the same type of trait: neuroticism (anxiety, harm avoidance), extroversion (sociability, other-centeredness), and openness to experience (risk taking, sensation seeking, curiosity, reward dependence), among others [53].

The concept of mood temperaments can be related to personality traits, the idea being that there is a biology to personality [54], even though there are also important obvious environmental and psychosocial components. Ever since Kraepelin [47], and then Kretschmer [55], almost a century ago, mild personality variants of mood illnesses have been described, using various terms for the same types: dysthymia (constitutional depression), hyperthymia (constitutional excitement), and cyclothymia (a combination of both). Recent genetic [56], biological [57], and clinical [58–60] studies indicate linkages between these mood temperaments and mood illnesses (unipolar depression and bipolar illness).

Thus, temperaments can be seen as dimensional extremes of normal personality traits which are biologically and genetically related to mood illnesses, being basically mild formes fruste of mood illnesses. In contrast, personality “disorders” are categorical constructs based mostly on psychoanalytic concepts, with little genetic basis [48] and limited biological knowledge about their etiologies or mechanisms [61]. If temperament concepts are valid, then the tendency to ignore them (hyperthymia is not even included in DSM revisions) would lead to an overdiagnosis of borderline personality.

It is worth emphasizing: temperament concepts are twice as old—about a hundred years—than the borderline concept, have stronger genetic and biological bases, and require no psychoanalytic ideological commitments.

Another area of spectrum overlap has to do with the concept of mixed states [45]: these agitated excited depressive states can easily be misdiagnosed as “MDD” with borderline personality, if one focuses on the irritability and impulsivity that are associated with them.

Mixed mood states are very important phenomenologically if it is true, as many studies indicate, that they are the most common type of mood episode [45, 62]. Kraepelin refused to base his nosology on polarity, presence or absence of depression versus mania, unlike DSM-III onwards, partly because he observed that most mood episodes in manic-depressive illness did not belong purely to one pole or the other [47, 63]. Most mood episodes were mixed, with perhaps the most common type of mixed mood episode being what is sometimes called agitated depression or more recently “mixed depression”: severe depression with psychomotor excitation, which involves marked mood reactivity, irritability, and usually psychomotor agitation [45]. Sometimes frank manic symptoms like brief periods of hyperactivity with flight of ideas and even heightened sexual drive and impulsivity occur. This mixed depression is completely ignored by the DSM nosology and labeled “major depression” as if all of these manic symptoms can be dismissed as nonexistent unless they happen for 4–7 days or longer [62]. Ignoring this unscientific, arbitrary DSM definition, recent data show that about one-half of all depressive episodes, whether occurring in bipolar disorder or not, are accompanied by multiple manic symptoms [64], as in the mixed depression description above.

Since DSM ignores the reality of these mixed mood episodes, it is likely that such patients often get mistakenly diagnosed with two conditions they do not have—MDD comorbid with borderline personality—rather than the one illness they do have: mixed depression (part of manic-depressive illness).

Given all of these diagnostic problems, how is one to differentiate a patient with borderline personality from a patient with bipolar illness, mood temperaments, or mixed depressive episodes?

We propose a simple solution, based on our prior work on the mistaken claim of bipolar overdiagnosis [65]:

Use non-symptom diagnostic validators to increase positive predictive value before applying DSM-style symptom definitions for mania or borderline personality.

All want to avoid false-positives. One could always avoid false-positives for low prevalence conditions (<20 % would be considered low prevalence, for statistical purposes), like bipolar illness, by just never diagnosing them. If bipolar illness happens in at most 5 % of the population, then if we never diagnose it, we would be right 95 % of the time. Many bipolar skeptics do just this: they have rarely seen a bipolar illness they did not refuse to diagnose. And they would be correct most of the time, statistically. But we presume we actually want to diagnose and help those suffering from this highly treatable and deadly illness.

So how can we avoid false positives while also diagnosing uncommon conditions frequently enough?

The DSM approach, loudly proclaimed by the leader of its 4th revision [15], is to attack any effort to broaden diagnostic definitions: spectrum concepts are rejected tout court and even seen as slightly evil. By making diagnoses narrow, we increase specificity to reduce false-positives. Hypomania can never be allowed to be defined as less than 4 days, even though there is zero evidence for that cutoff, and there is notable evidence for lower durations as meeting nosological validity requirements [22].

The problem with this approach, despite its use for three decades and counting, is that it can clearly be shown to be statistically doomed [65]. Predictive value is sensitive to prevalence. So with a low prevalence condition, like almost any psychiatric illness, if all we do is apply DSM criteria as our diagnostic test, we will always have a high rate of false-positives until we reach very high specificities of greater than 95 %, which is only attained with very sensitive laboratory tests in medicine, not with variable clinical judgments. To achieve a PPV of even 50 %, the rate that has been consistently misinterpreted in the Rhode Island study as representing overdiagnosis, we would need a diagnostic specificity for DSM criteria of 90 %, which is hardly attained with much more careful instruments like echocardiograms and x-rays. Clinical diagnostic reliability rarely reaches 80–90 % even in highly structured settings like research studies, much less clinical practice where it tends to be 50–70 % or lower [38, 65].

In other words, the low prevalence of a condition dooms any DSM-like effort to reduce false-positives by being diagnostically puritanical. There is only one solution: to increase the prevalence.

Of course we can’t go out into the community and infect people with bipolar illness, so how can we increase the prevalence?

We increase it by only allowing people into our clinical theater of evaluation who are already more likely to have bipolar illness [65]. We screen for clinical risk factors for bipolar illness other than the DSM criteria (i.e., mania): genetics, course, and treatment effects. Thus, if someone has bipolar genetics and early-onset depression before age 25 years, they already have a 50 % prior probability of meeting DSM-defined bipolar disorder criteria [65]. Then when we apply those criteria to that person, we will have 80 % positive and negative predictive values [65], much better than standard clinical practice and much better than even DSM-148 will ever be.

Now we can use this idea for the bipolar-borderline debate and see how available evidence applies.

Beginning with phenomenology, we found that manic mood symptoms differentiated bipolar illness from borderline personality. In one study by our group [66] in 260 patients with mood illnesses (68 % females), bipolar illness could be distinguished from borderline personality disorder, using DSM-IV definitions, based on a triad of euphoric mood (odds ratio, OR = 4.02, 95 % confidence intervals, CI:1.80, 9.15), mood episodicity (OR = 3.48; CI: 1.49–8.39), and increased goal-directed activities (OR = 3.9; CI: 1.73–8.96), whereas borderline personality disorder was not predicted by any mood symptoms examined. The only clinical feature predictive of borderline personality, as opposed to bipolar illness, was female gender (OR = 3.41; CI: 1.29–13.70).

In the BRIDGE study [67], which assessed an unselected mood population with depressive episodes, mixed depression as described above was assessed. Even using that very broad definition of mixed depression, borderline personality was able to be distinguished from mixed depression based on four of its DSM-IV features: fears of abandonment, identity disturbance, recurrent suicidal or self-mutilating behavior, and dissociative symptoms. The ability to distinguish these borderline features is all the more impressive when one appreciates that this is a sample of patients with clinical depression in which the prior probability of bipolar illness, simply based on presence of depression and the clinical/demographic features of the sample, is about 50 % [65].

Another key differentiating feature of phenomenology is parasuicidal self-harm in borderline personality [68]. A recent literature review of 51 articles [69] found that self-mutilation is common in borderline personality (50–80 % of cases) and is frequently repetitive (41 % of patients have more than 50 self-mutilation acts). In contrast, parasuicidal behavior is much less common in bipolar illness. In the National Comorbidity Survey (n = 5,877) the prevalence of self-harm among patients with type I bipolar illness was only 0.9 %. In other words, the difference between type I bipolar illness and borderline personality is about a 50–80-fold increased relative risk of parasuicidal self-harm in borderline personality. This is five- to eightfold higher than the association between tobacco and lung cancer [70]. A more definitive specific distinguishing feature between two clinical syndromes can hardly be found. The importance of the NCS data is that they are epidemiological, not clinical. They are based on determining prevalence of parasuicidal behavior in persons with bipolar illness who are in the community, not those who seek treatment in clinicians’ offices. By using clinically selected samples, higher rates of parasuicidal behavior are seen, even in bipolar type I illness, in some studies, but these samples involve a selection bias of those who seek help and do not generalize to the entire bipolar population. In contrast, the NCS study does generalize to the whole bipolar population and is probably the most valid data on which to base judgments about sexual trauma prevalence. Even with this limitation, the highest parasuicidal behavior rate reported in clinical studies is 36 %, which remains twofold less frequent than in borderline personality [71].

A critic could claim that matters are different for type II bipolar illness, as opposed to type I, despite the fact, as we’ve discussed above, that the assumption that there are major differences between type I and type II bipolar illness is not based on a solid scientific evidence base. Nonetheless, the parasuicidal behavior rate in type II bipolar illness has not been studied empirically, in our review of the literature. In contrast, suicide attempt and suicide rates have been demonstrated to be very similar between type I and type II bipolar illness [72].

Turning to the nosological validator of genetics, bipolar illness is one of the two most heritable mental illnesses, along with schizophrenia, both having about 80 % heritability, similar to Alzheimer’s dementia [48]. This rate is about twice as much as borderline personality or other personality traits or disorders, which tend to have about 40–50 % heritability [48, 73], a relatively low rate indicating that environmental aspects of etiology are as important as genes for borderline personality. In contrast, bipolar illness is almost completely genetic, similar to the heritability of height [73].

On the validator of course of illness, episodicity versus chronicity may not be definitive in differentiating bipolar illness from borderline personality if we include the concept of mood temperaments, which are by nature constant personality states. However, a key course feature that seems to differentiate these conditions is a history of sexual abuse. In a commonly cited meta-analysis of 21 studies, 50–76 % of patients with borderline personality disorder had experienced sexual trauma. Many have downplayed these results by claiming that the average effect size for the association between childhood sexual abuse and borderline personality is only moderate (r = 0.279) [74]. This attempt to downplay the high rate of sexual abuse in borderline personality, by focusing on the correlation coefficient, obfuscates a key distinction with bipolar illness, where sexual abuse occurs in less than 30 % of bipolar subjects [75, 76]. A recent systematic review including 3,407 bipolar patients found a 24 % prevalence of sexual trauma in bipolar illness [75], which is only slightly higher than the general population (13–17 % in women and 2.5–5 % in men: thus 15–22 % overall). The key finding is that sexual abuse is at least two times more common in borderline personality than in bipolar illness, and rates in bipolar illness are almost the same as the general population.

Regarding biological validators, bipolar illness is associated with enlargement of the amygdala and hippocampal atrophy, which is correlated with a number of mood episodes [77]. Borderline personality is not. Other biological changes found in borderline personality, such as deficits in integration between cognition and emotional processing stimuli [78], are not unique to borderline personality but are also found in other neuropsychiatric syndromes, including schizophrenia and bipolar illness [79].

On the final validator of treatment, about one-third of patients with bipolar illness are completely cured with mood stabilizers like lithium [28]. Patients with borderline personality have at best modest symptomatic improvement with mood stabilizers or neuroleptics and are not generally cured with medications [80]. In contrast, a substantial minority of patients with borderline personality appear to approve with psychotherapies, with complete remission in about one-third or more [81, 82], while bipolar illness never fully remits with psychotherapies alone [28].

In sum, bipolar illness can be differentiated from borderline personality with important diagnostic validators, as summarized in Table 6.1. It is irrelevant that many of these common predictors of these conditions, like sexual abuse and bipolar genetics and recurrent course, are not part of DSM criteria; they should be. Psychiatric diagnosis limited to DSM criteria is unscientific and poor psychiatric practice [30]. These other non-DSM predictors are important variables to increase the PPV of these conditions, thereby reducing false-positives in either direction.

Now we can apply these ideas to three case vignettes: one where the patient is hyperthymic not borderline, another borderline not bipolar, and a third bipolar only not bipolar plus borderline.