It is well known that comorbidity of bipolar disorder with substance use disorder is one of the most vexing problems in psychiatry. Although addiction comorbidity is elevated in all psychiatric illnesses, repeated epidemiological surveys have found that people with bipolar disorder (both types I and II) are more likely to have co-occurring substance use disorders over their lifetime than those with any other axis I disorder, with lifetime prevalence estimates ranging from 42% to 60% (1–3). The potential for substance use to disrupt the delivery of and adherence to care is high; because there are effective treatments for bipolar disorder, it is important that substance use disorders not interfere with access and adherence to care. Despite how common substance use disorders are in bipolar disorder, there is still no clear consensus on how treatment is best approached. The focus of this article is on strategies for diagnosis and treatment of patients with these complex disorders.

Diagnostic dilemmas

Accurately diagnosing bipolar disorder is difficult even under ideal circumstances; accurate diagnosis is made more difficult in the context of substance use. There is a temptation to make a cross-sectional diagnosis without corroboration with third parties, but this approach is inadequate for this group of patients. Bipolar disorder is a lifetime diagnosis, rather than a current diagnosis. Even a current manic episode does not secure a bipolar disorder diagnosis, and without a careful (and confirmed) history of prior episodes fully meeting DSM-IV criteria for manic, mixed, or hypomanic states, the diagnosis cannot be made. It is more difficult to be clear that these episodes were not substance-related when a retrospective diagnosis is made. There is some consensus, though, that although stimulants, cocaine, and phencyclidine intoxication can produce manic-like states, it is highly unlikely that these episodes will be directly caused by alcohol, sedative/hypnotic, cannabis, or opiate intoxication (4, 5). A manic episode can be reasonably diagnosed even in the face of such drug use, but hypomania and mixed states less so, although intoxication or withdrawal from any of these drugs can produce symptoms consistent with depression and anxiety disorders. Irritability, impulsive behavior, psychomotor agitation, and decreased sleep or need for sleep can be present in people who misuse substances due to intoxication or withdrawal, so clinicians must be precise in history taking so as not to mistake symptoms due to the direct effects of drug and alcohol use for symptoms of mania, hypomania, or mixed states. Major depressive episodes, required for the diagnosis of bipolar II (but not bipolar I) disorder, must not be due to the direct effects of substances or recovery from them, making retrospective diagnosis even more complicated as any drug intoxication or withdrawal may mimic symptoms of depression. It is difficult in taking a history to clearly differentiate mood episodes that occur independent of substances from those directly related to their effects.

In practical terms, inordinate numbers of patients seeking treatment for substance use disorders in whom bipolar disorder has been previously diagnosed do not, upon structured interview, actually have bipolar disorder; this has been found in inpatient, residential, and ambulatory treatment settings (6–9). Screening tools such as the Mood Disorder Questionnaire (MDQ) appear to be particularly bad for diagnosing bipolar disorder in patients with substance use disorders; one study found that fewer that one-quarter of substance-abusing patients in whom bipolar disorder was diagnosed by the MDQ had their diagnosis confirmed by structured interview (9). Perhaps this is the reason that Hall-Flavin et al. (6) found an unexpectedly low rate of bipolar disorder (5%) in patients in residential treatment for alcohol dependence, while finding unipolar major depressive disorder in 31% and anxiety disorders in 30% (6). Although this finding might in part be explained by decreased addiction treatment engagement by patients with bipolar disorder, it might also suggest that the high rates of substance abuse comorbidity found in clinical and, especially, epidemiological samples may be confounded by inaccurate overdiagnosis of bipolar disorder.

Impact on course

Although the diagnosis of bipolar disorder in patients who have substance use disorders may be difficult, patients who do have this comorbidity have a more complex course of illness. Patients with bipolar disorder and a current substance use disorder tend to have earlier onset of illness, more suicide attempts and completed suicides, worse health outcomes, poorer psychosocial functioning, and less adherence to treatment (10–25). At the same time, they tend to be younger and male (12). There is no certainty at all, however, that substance use is the cause of poor outcomes rather than the result of occurring in a more ill group, associated with earlier onset of illness (26).

The exception to the unclear causal relationship between substance use and outcome is in the case of cigarette smoking. Like other substance use disorders, nicotine dependence is more prevalent in patients with bipolar disorder than in the general population and has been clearly related to medical morbidity and early mortality, primarily due to cardiovascular disease (27–32). Although patients with bipolar disorder are more likely to smoke and smoking has been associated with more severe course of bipolar illness, including elevated risk for suicide attempts and suicidal thoughts, poor health outcomes are perhaps the most worrisome association (33–35). Whereas people may have the impression that suicide is the most common cause of premature death in bipolar disorder, far more people with bipolar disorder die prematurely from cardiovascular disease and malignancy than die from suicide (30).

When followed prospectively, patients with bipolar disorder and current substance use disorders may not have worse mood outcomes, especially when controlling for baseline factors such as age and age of onset of illness (36). Several studies have found that patients with current drug or alcohol use disorders are not less likely to recover from mood episodes, one of which had the surprising finding that the amount of alcohol subjects used was not related to whether or not they improved during treatment for depression (36, 37). In the Systematic Treatment Enhancement Program for Bipolar Disorder (STEP-BD), subjects with any substance use disorder (alcohol abuse or dependence or drug abuse or dependence) did not take longer to recover from index episodes of depression, with those with substance dependence, surprisingly, recovering somewhat more quickly than those with no substance use disorder. In another prospective study, the amount of alcohol that subjects consumed did not have any relationship with outcome (37). This is contrary to what has been the popular clinical wisdom: that it is necessary for patients with bipolar disorder to stop or decrease their use of substances to recover from mood episodes. Instead, it may be that patients with bipolar disorder and co-occurring substance use disorders have a more severe form of bipolar disorder to begin with and that the association is confounded by substance use (36). On the other hand, subjects with both past and current substance use disorders were more likely to “switch” to hypomania or mania while recovering from depression. Because this was the case whether subjects had a current or past substance use disorder, diagnosis suggests that this propensity to “switch” may be inherent in the patients rather than due to the direct effects of substance intoxication or withdrawal (36).

The question of “self-medication” of mood and other symptoms in bipolar disorder has never been adequately examined (38). Substance dependence (as opposed to abuse) is a diagnosis categorized primarily by loss of control over use; although some drug and alcohol use may be initiated in part because of a response by patients to their symptoms, it does not fully explain why someone loses control of their use. Because much drug and alcohol use in dependent subjects is cue-dependent and learned, it may instead be the case that specific mood symptoms cue patients to use drugs or alcohol and it is the reward associated with prior use that drives future use (39). Some evidence does suggest that the use of substances may be greater in depressive than in hypomanic states and differs based on mood state (40). If it were the case that drug and alcohol use vanished during periods of euthymia, however, self-medication might explain this, but this is not the case; instead drug and alcohol use in dependent patients may be better explained by the complex interaction between mood, anxiety, cue-based reward, and reward sensitivity (41).

Pharmacotherapy

The two primary questions in treatment of patients with bipolar disorder and cooccurring substance use disorders are the following: what treatments reduce the use of substances and what treatments for bipolar disorder are effective in this group? There are remarkably few published data to answer either question, and this is in part due to the difficulty in identifying and enrolling such subjects in trials (presumably those most affected are not seeking treatment) and in part due to greater focus on the part of pharmaceutical manufacturers (who invest the greatest amount of money in efficacy trials) on wider markets, i.e., general populations with bipolar disorder and not subgroups of patients. We can review available evidence.

It is best to exclude small, exploratory, and underpowered trials in discussions regarding the care of patients. These studies might inform further research, but the probability of there being false-positive findings with inadequate assessments of safety prohibit their discussion is such a setting. Despite this limitation, clinicians want to provide the most appropriate care for their patients, and because most efficacy trials exclude subjects with dual diagnoses, some guidance is still merited.

There are no replicated, adequately powered trials of any compounds in the treatment of substance use disorder in patients with bipolar disorder. The largest study of any treatment for this population was undertaken by AstraZeneca. Stedman et al. (42) studied quetiapine compared with placebo added to lithium or divalproex sodium in 362 subjects with co-occurring alcohol dependence and bipolar disorder, replicating a finding by Brown et al. (43). Despite quetiapine's having been repeatedly shown to be effective in multiple phases of bipolar disorder, these studies found no benefit for quetiapine for the primary drinking outcome and no clinically significant benefit for mood (44).

Maintenance treatment with sodium divalproex added to lithium for subjects with alcohol dependence and bipolar disorder decreased heavy drinking compared with placebo (45). This modestly sized randomized trial suggests that the addition of valproate may be of some benefit for such patients; although not replicated, the addition of valproate to lithium in patients with co-occurring disorders who have an inadequate response to lithium is consistent with current findings for bipolar disorder in general, so may be recommended (46). No other treatment specific to bipolar disorder has been shown to be effective in the treatment of any substance use disorder in patients with bipolar disorder.

Several treatments that have been shown to improve outcomes for substance use disorders have been studied in bipolar disorder, but none has unequivocally shown benefit. For alcohol dependence, there are two small studies of naltrexone and acamprosate in subjects with bipolar disorder, each of which finding a nonstatistically significant affect in favor of drug over placebo, but these studies were preliminary and underpowered (47, 48). Disulfiram, the other Food and Drug Administration (FDA)-approved treatment for alcohol dependence, has not been studied in bipolar disorder, and a trial of topiramate, a N-methyl-d-aspartate and glutamate modulator that has shown efficacy in decreasing heavy drinking in a nonpsychiatrically ill population, is currently under study (49, 50). No treatment for opiate dependence has been studied in bipolar disorder, and treatments for cannabis, stimulant, or cocaine dependence in bipolar disorder have not been studied in adequately powered, well-designed studies to be able to recommend them. (No treatment is FDA-approved for the treatment of cannabis, stimulant, or cocaine dependence.)

Despite this lack of clear data, it is reasonable to recommend that effective treatments for bipolar disorder should be used in all patients with bipolar disorder, regardless of comorbidity. The data suggesting that drug and alcohol use disorders in observational studies do not affect outcome for mood prospectively suggest that standard treatment is as effective for these patients and for patients without substance use disorder comorbidity; bipolar disorder is difficult to treat for the majority of patients in any case.

It is perhaps more difficult to determine whether to use substance use disorder-specific treatments in patients with bipolar disorder in the absence of evidence for their use. There is no reason to believe, for instance, that opiate substitution or nicotine replacement therapy would not be effective for patients with bipolar disorder, at least to some degree; the consequences of opiate and nicotine dependence on morbidity and mortality are undeniable, so withholding treatment for them because of a bipolar disorder diagnosis seems more risky rather than less. The standards of care for those disorders should be used first. Bupropion, although not effective for the treatment of bipolar depression, has not been shown to worsen mood outcomes in patients also taking an antimanic drug. Thus, although its efficacy in smoking cessation in bipolar disorder is not established, its safety may be assumed (51). For varenicline, however, decision-making is more difficult, because that drug is perceived to be associated with significant psychiatric side effects (although whether this is actually the case has not been established) (52). Because its safe use in bipolar disorder has not been established, clinicians should proceed with extreme caution if prescribing it.

As is also the case with anxiety disorders, the infrequent use of pharmacological treatments for addiction in patients with bipolar disorder is striking (53). There may be a perception that addictions are less severe in individuals with bipolar disorder than in the general population (this has not been established) and that substance use will improve with adequate treatment for mood. There is no evidence for this, and because in practice many patients with bipolar disorder remain symptomatic or have frequent relapses, this may be an unrealistic hope. Even in substance abuse treatment settings, the use of medical treatments for addictive disorders is still not widespread (54). However, the health consequences of continued abuse of substances may merit their use even though specific evidence of efficacy in bipolar disorder is lacking.

Psychosocial treatment

Treatment for substance use disorders is primarily effective in the context of concurrent psychosocial treatment, so it would be expected that this would also be the case in co-occurring bipolar disorder. One question is whether standard psychosocial treatments for substance use disorders are effective in co-occurring disorders. In an ambitious study, Weiss et al. (55) found that group therapy specifically modified to address substance use disorders and bipolar disorder concurrently produced improved substance use outcomes compared with the randomly assigned comparison treatment, group therapy for addiction. Mood outcomes were not improved in the bipolar disorder treatment group, consistent with observational findings. This treatment has also been studied in a modified form intended to facilitate its use in community settings (56).

A combination of motivational interviewing and modified cognitive behavior therapy has shown promise in this group (57). Motivational interviewing and other strategies for working with patients with addictions, such as 12-step facilitation, should be a part of every psychiatric treater's approach. Patients with bipolar disorder and addictions appear to have a longer delay to treatment and appear to be underrepresented in residential treatment settings, so any office-based interventions to facilitate behavioral change may be beneficial.

Overdiagnosis of comorbidity

That there is a high rate of substance use disorder comorbidity in bipolar disorder is not in dispute; whether the rates are as high as current data suggest is uncertain, because there is almost certainly overdiagnosis of bipolar disorder in people who misuse substances due to symptom overlap, with symptoms associated with substance use and recovering from its effects (irritability and anger, psychomotor agitation, impulsive behavior, and decreased sleep mimicking mania and hypomania) probably increasing the rates of bipolar disorder in epidemiological samples. The high rates of substance use disorders (other than smoking) in clinical and research samples tend to be due to the retrospective diagnosis of past substance use disorders, because the rate of current substance use disorders is quite low (5% in a Stanley Foundation study of bipolar disorder, 12% in the STEP-BD) (11, 12). The rate of current substance use disorders in these studies is much lower than that found in a large study of major depressive disorder, Sequenced Treatment Alternatives to Relieve Depression (STAR*D), in which nearly 30% of subjects enrolled had a concurrent drug or alcohol problem (58). Because comorbid disorders tend to be more likely to be overrepresented in treatment samples, as in Berkson's bias, it is possible that the rate of actual comorbidity is lower that what was found in STEP-BD. Many would say that the low rates of current substance use disorders in treatment-seeking patients with bipolar disorder is due to avoidance of care by those currently using substances, but it is not clear why this would be the case in STEP-BD and not for unipolar depressed subjects in STAR*D.

Change in focus from mood to health outcomes

A focus on mood outcomes and their consequences, such as suicide, has dominated the approach to the study of co-occurring bipolar and substance use disorders, but it is perhaps time to shift the focus from one on mood to one concerned with health outcomes for patients. As we become more aware of the poor health outcomes for patients with bipolar disorder (even accounting for death by suicide, the risk of death from medical causes is elevated across the lifespan in patients with bipolar disorder, perhaps double that of patients without mental illness), we should modify our practices to be able to change these outcomes. By increasing appropriate diagnosis of bipolar disorder and by reducing the stigma associated with it, we might be able to improve mood outcomes for more people, but our ability to improve mood outcomes for patients already receiving appropriate treatment for bipolar disorder is probably quite limited. By integrating treatment to change health risk behaviors, especially smoking but also other drug and alcohol misuse and other lifestyle practices, into standard care, there is the possibility that our care will decrease medical morbidity and mortality. This will require appropriate training for psychiatric treaters in strategies for behavior modification and in treatments for specific addictions, while improving the awareness of the health risks of the medications that we use to treat bipolar disorder.

There are three primary recommendations to be made from this clinical summary:

• The diagnosis of bipolar disorder with co-occurring substance use disorders must be made carefully in clinical practice . This may seem self-evident, but there is certainly overdiagnosis of bipolar disorder in patients in addiction treatment settings, and exposing these patients to potentially harmful treatments for bipolar disorder should be avoided. Because the diagnosis of bipolar disorder is frequently made retrospectively, potentially confounding symptoms should be carefully evaluated to avoid overdiagnosis, and reliance on third-party information is essential. Conversely, the diagnosis of substance use disorders should be carefully made in patients with bipolar disorder, because the stigma of drug and alcohol use may often skew reporting.

• Substance use disorders in patients with bipolar disorder should be specifically treated . Despite limited evidence, there is no clear reason to avoid the use of pharmacological treatment for addiction, especially for nicotine, alcohol, and opiates, and psychiatric treaters should be prepared to use currently available medical and psychological treatments for addictions.

• Substance misuse in patients with bipolar disorder contributes to excess morbidity and mortality . Most evidence suggests that although patients with bipolar disorder and substance use disorders are more ill, it is not clear in longitudinal studies that they are less responsive to treatment for mood episodes. The focus should shift to improving health outcomes. This requires that clinicians be able to facilitate and monitor behavioral change and should maintain training to do so.