The results are limited by the use of fully structured lay-administered CIDI interviews rather than clinician-administered interviews, although the clinical reappraisal study found good concordance of CIDI diagnoses with blinded clinical diagnoses based on the SCID, reducing concern about validity. Concordance was quite high for BP-I (κ = .88) and any BPD (κ = .94), but considerably lower, although still acceptable, for BP-II (κ = .50) and sub-threshold BPD (κ = .51), due to the CIDI having difficulty distinguishing between BP-II and sub-threshold BPD.22
A related limitation is absence of information on mixed episodes, rapid cycling or brief episodes that could be assessed in more flexible semi-structured clinical interviews. Our definition of sub-threshold BPD is consequently more restrictive than the definitions proposed by clinical researchers.16, 35-37
The less flexible assessment than clinical interviews also could have led to overestimation of comorbidity and bias in retrospective recall of persistence. No data are available on accuracy of these reports. The less flexible nature of the CIDI than clinical interviews also could have led to over-estimation of comorbidity and bias in estimated clinical severity and persistence. No data were collected on accuracy of these reports.
Within the context of these limitations, the results provide the first nationally representative US general population prevalence estimates of sub-threshold bipolar disorder. BP-I and BP-II prevalence estimates (1.0-1.1%) are consistent with estimates from earlier population-based studies,1, 3-8
with the exception of a much higher lifetime prevalence estimate of BP-I (3.3%) in a very large recent national survey of the US.2
No clinical validation of the latter estimate was reported. It is noteworthy that the NCS-R clinical reappraisal study confirmed the much lower NCS-R BP-I prevalence estimate. Estimated averages of 77.6 lifetime episodes for BP-I and 63.6 for BP-II are somewhat higher than in prospective studies and family studies,38, 39
indicating possible over-estimation due to retrospective recall bias.
For reasons noted above regarding limitations, the NCS-R sub-threshold BPD prevalence estimate is likely to be a lower bound, although it is broadly consistent with two large community epidemiological surveys in Europe.40, 41
The NCS-R results clearly document the clinical significance of sub-threshold BPD, as the vast majority of sub-threshold cases had moderate-severe symptom profiles and role impairment based on standard rating scales. As one might expect, there was lower episode persistence and a lower severe-to-moderate ratio among sub-threshold vs. threshold cases. Consistent with previous research, the proportions of depressive episodes rated severe were higher for BP-II than BP-I and lowest for sub-threshold BPD.15
The more striking results from the perspective of sub-threshold BPD, though, are that the clinical severity, role impairment, and comorbidity of sub-threshold BPD are all quite high and, indeed, comparable to those of non-bipolar major depression reported in previous NCS-R analyses.42
These findings strongly argue for the clinical significance of sub-threshold BPD.
After controlling for time at risk,43
the high comorbidity of threshold BPD is consistent with prior clinical44, 45
and population-based2, 3, 46-48
studies, although comorbidity with substance use disorders was more prominently featured in previous studies. The higher comorbidity found here with anxiety and impulse-control disorders was less consistently studied in previous research.49, 50
Despite the higher disorder-specific comorbidity of threshold than sub-threshold BPD, comorbidity with at least one other disorder was nearly as common in sub-threshold (88.4%) as threshold (95.8-97.7%) cases. This means that the generally lower bivariate comorbidity of sub-threshold than threshold BPD is due to lower multimorbidity51
(i.e., comorbidity with multiple conditions). This pervasive comorbidity across the bipolar spectrum is suggestive of disturbances in multiple regulatory systems and should be a topic for future research.
The vast majority of NCS-R respondents with threshold (87.1-91.5%) or sub-threshold (67.8%) BPD reported lifetime treatment for emotional problems. However, treatment in the year before interview was lower both for threshold (67.3-65.8%) and sub-threshold (36.7%) cases and only a minority received appropriate medication (25.0% BP-I, 15.4% BP-III, 8.1% sub-threshold BPD). Appropriate maintenance medication of asymptomatic lifetime cases was even lower (17.9% BP-I, 15.6% BP-III, 3.2% sub-threshold BPD). The proportions receiving inappropriate medication (primarily antidepressants in the absence of antimania agents) were considerably higher (31.4% 12-month cases, 25.1% asymptomatic lifetime cases), especially among patients in general medical treatment (73.1% 12-month cases, 65.6% asymptomatic lifetime cases). Appropriate medication was much higher and inappropriate medication lower among patients in psychiatric treatment, although fewer than half of psychiatric patients took appropriate medications (45.0% of 12-month cases, 41.1% of asymptomatic lifetime cases). The high use of inappropriate medications is a concern given the dangers associated with use of antidepressants in the absence of mood stabilizers to treat BPD.52
It is noteworthy that the treatment percentages represent patients who took
medications. The numbers who were prescribed
but did not take medications were not recorded in the survey. It is quite possible that higher proportions were prescribed antimanic agents but did not take them, as subjective distress is greater for depression than mania/hypomania.11
Although providing only a lower bound estimate on prevalence of sub-threshold BPD, our results reinforce the argument of others that clinically significant sub-threshold BPD is at least as common as threshold BPD.16, 35-37
Although most of those with BPD receive treatment due to comorbid disorders, the lack of recognition of their underlying bipolarity leads to only a minority receiving appropriate treatment. Clearly, more comprehensive screening of bipolar symptoms is needed among patients presenting for treatment of other Axis I disorders. The failure to recognize sub-threshold BPD can also reduce the precision of estimates and lead to bias in genetic and other etiologic studies of mood disorders.53-55
Additional research is needed to resolve uncertainty regarding the most appropriate boundary distinctions for BPD. This uncertainty remains a major impediment to advancing our understanding of the bipolar spectrum in the population.