The most common incident disorders in this study were MDD, alcohol abuse and dependence, and GAD. The incidence of MDD was 1.51, virtually identical to the rate in the ECA (1.52),11
but lower than the rates in Edmonton (2.79),12
the Netherlands (2.72),13
and Finland (2.05),15
and higher than those observed in the older long-term follow-up studies (0.24–0.45)1–3,5
The incidence of panic disorder (0.62) was similar to rates in the ECA (0.56) 11, 19, 20
and the Netherlands (0.78)13
surveys. The incidence of social phobia (0.32) was comparable to the rate in the Baltimore Follow-Up Survey (0.45)11
but lower than those in the ECA (0.94)11
and the Netherlands (0.93).13
One-year incidence of alcohol dependence was 1.70, higher than was observed in the Baltimore (0.46)23
or the Netherlands (0.49)13
surveys. This study also found a lower incidence of alcohol abuse (1.02) than in the Netherlands survey (2.38). Incidence rates for drug abuse and dependence in this study (0.28, 0.32) and others (0.28, 0.27)13
were low. The lower rates reported for most disorders in the long-term studies could be attributed, in part, to smaller numbers of incident cases and the inevitable impact of attrition. Discrepancies in incidence rates between surveys may also reflect differences in survey design, and/or environmental or genetic factors. Differences may also be due to diagnostic criteria used in the current study (i.e., DSM-IV) and prior studies that used earlier DSM classifications or ICD-10 criteria.
Consistent with most prior cross-sectional27, 50–54, 65–71
and longitudinal research,2,11–14
incidence rates of MDD and anxiety disorders except social phobia were greater among women, while incidence rates of most substance use disorders were greater among men. Also consistent with these prevalence surveys, there were no sex differences in the incidence of bipolar I and II disorders. However, unlike earlier prospective studies,11–13
this study found inverse relationships of all assessed disorders with age. Although cross-sectional studies17, 18, 27, 50–54
have consistently reported inverse relationships of most disorders with age, it remained unclear whether these associations were real, artifactual due to longer duration of illness, or due to mortality, recall, or other biases. The findings on age derived from prospectively determined incidence rates strongly suggest that the observed age differentials represent true differences in first incidence, with greater incidence among younger cohorts.
This study also identified other sociodemographic risk factors for DSM-IV disorders not generally reported in prior research due to limitations in sample size. Incidences of alcohol and drug abuse and dependence, MDD, and GAD were greater among separated/divorced/widowed individuals, a result that extended to never-married individuals for alcohol abuse and dependence. While these findings do not entirely clarify the causal relationship between marital status and psychopathology, they indicate that the relationship is not due solely to unmarried status resulting from preexisting psychopathology. Also, low family income was significantly related to risks of most anxiety disorders and MDD, but not other mood or substance use disorders. Incident bipolar I disorder, however, was associated with less than a high school education. Taken together, these results highlight age as an important general risk factor for DSM-IV substance use, mood, and anxiety disorders, whereas effects of sex and lower socioeconomic status appear to be disorder-specific. Future analyses of the NESARC data will test whether other prospectively assessed risk factors for psychiatric disorders are disorder-specific to help define the boundaries of DSM-IV disorders.
Little has been reported about incidence rates among race-ethnic minorities. In contrast to the 1 prior study,72
this study did not find elevated rates among Hispanics. However, the risk of incident GAD was lower among Hispanics and the risk of alcohol abuse was lower among Blacks. Future research is needed to explain the lower risk of these disorders among minorities.
To our knowledge, this is the first study to examine psychopathologic predictors of a broad range of incident DSM-IV disorders in a national sample and determine whether disorder-specific associations reflected common or unique factors. Some associations between psychopathologic predictors of incident disorders remained statistically significant, although reduced in magnitude, once baseline comorbidity was controlled. The drops in magnitude suggests common causal factors underlying the disorder-specific associations. However, the remaining significance of these associations suggests unique factors driving disorder-specific associations.
Consistent with the ECA data,72
baseline dysthymia predicted incident MDD. Also not surprisingly, baseline MDD predicted incident bipolar II disorder, suggesting that MDD occurs prior to hypomania in the development of bipolar II disorder. Consistent with prospective studies,28,29
alcohol abuse and dependence showed strong reciprocal temporal relationships but drug abuse only predicted incident drug dependence. The reciprocal relationship between alcohol abuse and dependence suggests that strong common factors may underlie the comorbid relationship and additionally provides support for elimination of the hierarchy between alcohol abuse and dependence in future DSM revisions. Further research is needed, however, on specific drug use disorders to support elimination of the abuse-dependence hierarchy in the DSM-IV in view of the unidirectional relationship between drug abuse and dependence observed in this study.
In general, baseline anxiety disorders more often predicted other incident anxiety disorders than mood disorders. Panic disorder predicted incident social and specific phobias and GAD, GAD predicted incident panic disorder, social phobia predicted incident GAD, and PTSD predicted incident panic disorder and specific phobia. Baseline social phobia predicted incident GAD. In only 3 instances did a mood disorder predict an incident anxiety disorder: bipolar I predicted both incident panic disorder and GAD, and MDD predicted incident GAD. These results are broadly consistent with most,73–79
but not all,80, 81
longitudinal studies showing that onsets of anxiety disorders are more often followed than preceded by the onset of depressive disorders. The observed temporal relationships, especially among anxiety disorders, may also reflect overlap of core DSM-IV symptoms among these disorders.
One of the most interesting findings of this study was the reciprocal temporal relationships between MDD and GAD, and GAD and panic disorder. These findings suggest the existence of strong common causes underlying those disorders, stronger than those common factors characterizing comorbidity among other disorders assessed in this study except for alcohol abuse and dependence. The observed reciprocal relationship between MDD and GAD is consistent with results of twin studies showing these disorders to share joint genetic susceptibility.82–86
Findings on the relationship between GAD and panic disorder show GAD to be etiologically distinct from panic disorder,86
but more recent studies support a shared diathesis between GAD and panic disorder87
or additive genetic influences common to GAD and panic disorder in the presence of a nonadditive genetic contribution specific to panic disorder.88
The present results suggest that genetic research be expanded to encompass MDD, GAD, and panic disorder, along with other mood and anxiety disorders, for the purpose of unraveling common and unique genetic and environmental influences underlying comorbidity.
By definition, PDs constitute enduring patterns of inner experiences and behaviors that are pervasive, inflexible and stable over time, with onsets in adolescence or early adulthood,1
and highly comorbid with mood and anxiety disorders.17, 18, 27, 50–54
Therefore, it was not surprising that PDs predicted these incident disorders. Borderline and schizotypal PDs predicted incident MDD, bipolar I, panic disorder, GAD, and social phobia. Borderline PD also predicted incident specific phobia, and narcissistic PD predicted incident bipolar I and GAD. That PDs predicted many mood and anxiety disorders tempts speculation that genetic risk shared among anxiety and mood disorders85,88
might be mediated by PDs, especially borderline and schizotypal PDs or traits.
Longitudinal and twin studies89–96
have consistently found antisocial behavior or conduct disorder in childhood or late adolescence to predict alcohol dependence in early adulthood. By contrast, antisocial PD did not predict incident alcohol or drug use disorders in this study. This discrepancy suggests that the relationship between childhood antisocial behavior and later substance use disorders may not be consistent across developmental stages.97
Adolescence and early adulthood are periods associated with the highest prevalence of substance use and the relationship measured at these stages of life may be different from what would be observed in later adulthood. Thus, this linkage between antisocial behaviors and substance use disorders may be evident among younger individuals not captured in the NESARC sample. However, antisocial behavior in prior studies may be predicting early-onset substance use disorders, leaving open the possibility that other personality psychopathology, such as borderline, narcissistic, or schizotypal PD or traits, could influence the development of later-onset substance use disorders as observed in this study. Future longitudinal research should be extended to adolescence and later adulthood and incorporate measures of a broad spectrum of personality psychopathology, with particular focus on sex and age differences in the manifestations of externalizing and internalizing psychopathology predictive of substance use disorders.
Another significant finding is that substance use disorders did not predict any incident mood or anxiety disorder. By contrast, baseline bipolar I predicted incident drug abuse, and baseline panic disorder predicted incident drug dependence. These results are consistent with recent evidence from a twin study showing that the risk of alcohol dependence was substantially increased by a prior episode of MDD, but a previous episode of alcohol dependence did not increase the risk of MDD.98
Although these results may be consistent with the self-medication hypothesis, other mechanisms such as shared underlying liability arising from the same genetic or environmental risk factors cannot be excluded.
Limitations of this study are noted. Although this study represents the largest follow-up survey of psychiatric disorders conducted to date, future prospective research with longer follow-up periods and those incorporating clinical interviews and collateral reports are also indicated. Since attrition between the Wave 1 and Wave 2 NESARC was small (13.3%) and the Wave 2 data were adjusted for nonresponse due to sociodemographic characteristics and presence of any substance use or other psychiatric disorder at Wave 1, attrition is not likely to have had as substantial effect on the incidence rates and risk associations examined in this study. Although the NESARC survey design included group quarters, some special populations, such as those under 18 or respondents in jail or hospitalized at the time of the interview, were not included in the sample. Finally, this study assessed DSM-IV disorders categorically, in conformity with clinical tradition. It is acknowledged that a dimensional approach to the measurement of DSM-IV disorders may have great merit for understanding the pathophysiology of each disorder and the comorbidity it shares with others.
In summary, this study has increased our knowledge of sociodemographic and psychopathologic risk factors for major DSM-IV substance use, mood, and anxiety disorders. The greater incidence of all these disorders in the youngest cohort underscores the need for heightened vigilance in identifying and treating such disorders among young adults. This study also provides a framework for future analyses focusing prospectively on other risk factors for the incidence, remission, and recurrence of specific disorders. Taken together, the findings of this study call for more research in the rapidly growing field of psychiatric genetics that has begun to expand phenotypic definitions beyond the study of a single disorder or trait to a range of phenotypes that show a high degree of comorbidity. Work in this area is beginning to identify latent genetic risk factors that indicate shared genetic susceptibility across a range of diagnostic phenotypes.99–104
Information on sociodemographic and psychopathologic risk factors prospectively identified in this study may also begin to inform a new class of preventive interventions aimed at preventing comorbidity (i.e., the prevention of the first onset of a second or set of disorders). With regard to clinical implications, clearer data about the risks of future disorders posed by chronologically primary disorders can increase efficiency of treatment planning and provide important information to patients at risk of developing secondary disorders. Primary prevention of secondary disorders would be feasible even when the comorbid conditions share common causes. The onset of the secondary disorder is not inevitable because common causes often have modifiable mediators.