Affective symptomatic recovery after a first manic episode in patients with bipolar disorder and no cannabis use disorders (no cannabis), cannabis use disorders that began after the onset of bipolar disorder (bipolar first), and cannabis use disorders that began before the onset of bipolar disorder (cannabis first). *Censored data (patients dropping out of the study before recovery). The cannabis first group survival curve significantly differs from those of the other groups (log-rank χ22 = 11.8; P = .003).
Affective symptomatic recurrence after a first manic episode in patients with bipolar disorder and no cannabis use disorders (no cannabis), cannabis use disorders that began after the onset of bipolar disorder (bipolar first), and cannabis use disorders that began before the onset of bipolar disorder (cannabis first). *Censored data (patients dropping out of the study before recurrence). The survival curves are not significantly different (P = .16).
Survival curves illustrating recovery from a cannabis use disorder after a first manic episode in patients with bipolar disorder and a cannabis use disorder that began after the onset of the bipolar disorder (bipolar first) and a cannabis use disorder that began before the onset of bipolar disorder (cannabis first). *Censored data (patients dropping out of the study before recovery). The survival curves are significantly different (log-rank χ21 = 13.7; P<.001).
Strakowski SM, DelBello MP, Fleck DE, Adler CM, Anthenelli RM, Keck PE, Arnold LM, Amicone J. Effects of Co-occurring Cannabis Use Disorders on the Course of Bipolar Disorder After a First Hospitalization for Mania. Arch Gen Psychiatry. 2007;64(1):57-64. doi:10.1001/archpsyc.64.1.57
Copyright 2007 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.2007
Cannabis use disorders commonly co-occur in bipolar disorder; however, the effects of cannabis abuse on outcome have been minimally studied.
To identify how the sequence of the onsets of a cannabis use disorder and bipolar disorder is associated with the subsequent course of each condition.
Academic medical center.
Patients (N = 144) were studied who met criteria for bipolar I disorder (manic or mixed), were 12 to 45 years old, and had no previous hospitalizations and minimal previous treatment. Patients were followed up for up to 5 years and included 33 in whom the onset of a cannabis use disorder preceded the onset of bipolar disorder (cannabis first), 36 in whom bipolar disorder onset preceded the onset of cannabis abuse (bipolar first), and 75 with bipolar disorder only.
Main Outcome Measures
Symptomatic recovery and recurrence of both conditions and percentage of follow-up time with affective and cannabis use disorder symptoms.
The cannabis first group exhibited better recovery than the other groups, although when adjusted for potential mediator variables these results did not persist. Cannabis use was associated with more time in affective episodes and with rapid cycling. Most cannabis use disorders remitted immediately after hospitalization, followed by rapid rates of recurrence.
The effects of the sequence of onsets of bipolar and cannabis use disorders were less pronounced than observed in co-occurring alcohol and bipolar disorders. Aggressive drug abuse treatment immediately after a first psychiatric hospitalization might decrease rates of recurrence and new cases of cannabis use disorder in the course of bipolar disorder.
People with bipolar disorder frequently struggle with substance abuse and dependence.1- 11 For example, the Epidemiologic Catchment Area Study found that 46% of patients with bipolar disorder exhibited a lifetime history of an alcohol use disorder and that 41% had a history of other drug abuse or dependence.1 Typically, cannabis is the most commonly abused drug in individuals with bipolar disorder, mirroring the relative rates of drugs of abuse in the general population.2,3,6,9,11- 13 In some studies1,3,7,13 of bipolar disorder, particularly those with younger patients, the rates of cannabis use disorders equal or exceed those of alcohol abuse or dependence. There are several large, prospective investigations8,13 of the effects of alcohol use disorders on the course of bipolar illness. In contrast, the effects of cannabis have been much less studied. Some investigators14,15 have implied that cannabis may actually be mood stabilizing in patients with bipolar disorder, although this suggestion lacks research support. The few systematic studies published suggest that drug abuse in general and cannabis abuse in particular are associated with poor treatment adherence,4,16 increased duration or severity of mania,3,11 and negative outcome in bipolar disorder.3- 7,9- 11 Moreover, cannabis abuse complicated the treatment of co-occurring alcohol use disorders in 1 study of patients with bipolar disorder.17 However, the relationships between cannabis use and bipolar disorders are complex and remain incompletely described.3
In studies of co-occurring alcohol use and bipolar disorders, investigators have attempted to address these complexities by using the sequence of onset of the 2 conditions to define separate patient groups.8,13 These studies found that, when the onset of bipolar disorder occurs before the onset of alcoholism, patients have a poorer outcome (eg, slower recovery, earlier onset, and more symptoms during follow-up) than when the alcohol use disorder occurs first. From these results, investigators suggested that a subgroup of patients with bipolar disorder develop a relatively milder form of affective illness that is expressed only after extended exposure to alcohol abuse. Whether drugs other than alcohol, including cannabis, exhibit similar associations is not yet known.
With these considerations in mind, we prospectively studied the courses of bipolar and cannabis use disorders after a first psychiatric hospitalization for mania. We compared the courses of illness between patients whose bipolar onset preceded the development of a cannabis use disorder and those in whom the cannabis use disorder was antecedent. First-episode patients were studied to improve the validity of age-at-onset estimates and to minimize the effects of long-term illness progression and chronicity, which might weaken associations between the disorders.3,4,6,7,11,13 The aim of this analysis is to address the hypothesis that the relative sequence of cannabis use and bipolar disorders onset affects associations between these conditions as reflected in the courses of illness of both. Specifically, we predicted that, when the onset of bipolar disorder followed the onset of cannabis abuse, the course of both the bipolar and cannabis use disorders would be better than when the opposite sequence occurred, similar to what has been reported with alcohol. In contrast, we predicted that cannabis abuse in general would be more strongly linked to mania than to depression.
The methods used for this study have been previously published13 and so will be described relatively briefly, focusing on aspects relevant to the present analysis.
Patients with bipolar disorder (N = 144) were recruited as part of the University of Cincinnati First-Episode Mania Study.3,13,18,19 Inclusion criteria included (1) met the DSM-IV criteria for bipolar disorder, manic or mixed (bipolar I disorder), with a Young Mania Rating Scale20 score greater than 20, (2) age 12 to 45 years, (3) no previous psychiatric hospitalizations, (4) less than 1 month of cumulative lifetime previous thymoleptic or antipsychotic drug exposure, (5) English speaking, and (6) able to return for follow-up visits. Patients were excluded (1) if psychiatric symptoms were due entirely to acute medical illness or acute drug or alcohol withdrawal or (2) if patients had an identified low IQ (<70). Written informed consent was obtained from adult patients and from a parent or guardian of adolescent patients (with the patient's assent) after the details of the study were explained. The institutional review boards of both the University of Cincinnati College of Medicine and the Cincinnati Children's Hospital Medical Center approved this protocol.
Patients were recruited from consecutive hospitalizations between June 1, 1996, and October 1, 2003. Approximately 95% of the eligible patients (n = 172) signed the informed consent form, and 144 of these (83.7%) had at least 4 months of follow-up and are included in this study.13 Four months was chosen as the minimum follow-up period to be consistent with previous work.3,13,18,19 There were no meaningful differences in most clinical and demographic variables among those who did and did not achieve this milestone.13 Eighteen (64%) of the 28 patients who dropped out of the study had a history of a cannabis use disorder, which was not significantly different than the rate observed in the study participants (48%; χ22 = 2.5; P = .11). In addition, these 18 patients were evenly distributed between the 2 comorbid groups (9 in the bipolar first group and 9 in the cannabis first group).
The diagnosis of DSM-IV bipolar disorder was established using the Structured Clinical Interview for DSM-IV Axis I Disorders–Patient Edition (SCID-I/P)21 administered by trained and reliable research clinicians.13 Symptoms were assessed using the Young Mania Rating Scale20 and the 17-item Hamilton Depression Rating Scale.22 The presence or absence of psychosis was determined using the SCID-I/P and the Scale for the Assessment of Positive Symptoms.23 Index symptoms were rated for the worst period during the current episode, typically coinciding with hospital admission. The investigators established good interrater reliability for these measures.13
Trained research assistants assessed substance use disorders using the Substance Use Disorders module of the SCID-I/P and the Addiction Severity Index (ASI),24 again with good interrater reliability.13 Drug and alcohol use in the 30 days before hospital admission was rated using the ASI. An ASI overall severity-of-use score was determined for drug and alcohol use based on a 9-point scale (0-1, no real problem; 2-3, slight problem, substance abuse treatment probably not necessary; 4-5, moderate problem, some treatment indicated; 6-7, considerable problem, treatment necessary; and 8-9, extreme problem, treatment absolutely necessary). Demographic information (age, sex, ethnicity, and years of education) was obtained from patient interviews and medical records.
The age at onset of bipolar disorder was defined as the age at which the first DSM-IV affective episode began, established using the SCID-I/P. The index manic or mixed episode was the first affective episode in 89 of these patients (61.8%). Forty-five patients (31.2%) experienced 1 (n = 20) or more (n = 25) previous depressive episodes, and 23 (16.0%) had 1 (n = 14) or more (n = 9) previous untreated hypomanic or mild manic episodes. For all the patients, this was the first manic or mixed episode requiring hospitalization. The age at onset of cannabis use disorders was defined as the age at which patients first met the criteria for cannabis abuse. This was determined by integrating information from the SCID-I/P and the ASI. Interrater reliability for age-at-onset estimates was high (intraclass correlation coefficient, >0.90).
Patients were classified into 3 groups by age at onset. The no cannabis group consisted of patients with bipolar disorder with no history of cannabis use disorders (n = 75; 52.1% of the sample). The bipolar first group included patients in whom the onset of bipolar disorder occurred before or concurrently with the onset of the cannabis use disorder (n = 36; 25.0%). “Concurrently with” was defined as within 1 year of the onset of bipolar disorder.13 Finally, in the cannabis first group, the onset of the cannabis use disorder preceded the onset of bipolar disorder by more than 1 year (n = 33; 22.9%). Rates of completing at least 1 year of follow-up were 75% (56 patients) for the no cannabis group, 83% (30 patients) for the bipolar first group, and 73% (24 patients) for the cannabis first group (χ22 = 2.1; P = .35). The primary reasons for patients not reaching the 1-year landmark typically were that patients moved from the area or were lost to follow-up.
After hospital discharge, patients were reevaluated at 1 and 4 months and then every 4 months for up to 8 years. The present analysis was restricted to a maximum of 5 years of follow-up because the number of available patients after that time was relatively small. The mean (SD) follow-up for the entire sample was 135 (89) weeks (2.6 years). Twenty-five percent of the sample had more than 4 years of follow-up, and half had more than 2.5 years of follow-up.
The general study design is based on previous work.4,8,13,25 At each follow-up visit, investigators reviewed the previous interval, week-by-week, for affective symptoms and substance use. Particular attention was given to periods of symptom changes, and calendar methods were used to identify changing drug and alcohol use patterns. Each follow-up review included every item of the symptom rating scales, the relevant modules of the SCID-I/P, the ASI, and ratings of the severity of affective and cannabis use disorder syndromes. For both the bipolar and cannabis use disorders, 6-point week-by-week ratings of symptom severity were determined from these assessments (Table 1).13 For alcohol and cannabis use disorders, calendar methods were used to identify change points and use, but only the ASI overall rating coupled with the SCID-I/P module were used to make determinations of the week-by-week severity ratings. From these ratings, periods of recovery and recurrence were identified, and the percentage of weeks in different phases of illness could be calculated. Urine samples for toxicology screening were not routinely obtained. Independent raters, blind to each other's scores, obtained the substance use and affective symptom ratings.
Affective symptomatic recovery from bipolar disorder was defined as at least 8 contiguous weeks with symptom severity ratings of 1 or 2 (Table 1).13 Affective (symptomatic) recurrence was defined as at least 1 week of several new significant subsyndromal symptoms (scores >3) (Table 1). Recovery from cannabis use disorders was similarly defined as at least 8 contiguous weeks with symptom severity ratings of 1 or 2 (Table 1). Recurrence of cannabis use disorders was defined as at least 1 week of new symptom severity scores greater than or equal to 3 (Table 1). Other drugs of abuse were relatively uncommon in this sample (<10%) and were always accompanied by cannabis use3,13 and so were coded as present or absent and covaried in analyses when indicated. Finally, rapid cycling was defined as 4 or more affective episodes in any 52-week period; new episodes were defined by a phase shift, with or without remission, or any new episode after recovery.
Because this was a naturalistic study, investigators did not administer treatment. Treatments that patients received were reviewed, and adherence was recorded as (1) full adherence, in which the medication was taken more than 75% of the time as prescribed; (2) total nonadherence, in which the medication was taken less than 25% of the time as prescribed; and (3) partial nonadherence, in which the medication was taken between these 2 extremes.4,13,25 This rating was obtained by reviewing week-by-week interval medication use with patients and with family members or clinicians when necessary. Serum levels were not obtained. From this review, the percentage of follow-up in which patients exhibited full adherence to at least 1 putative mood stabilizer was determined for each medication, and a mean (total adherence) score across medications was obtained. Putative mood stabilizers were broadly defined and included lithium, divalproex, carbamazepine, lamotrigine, topiramate, and atypical antipsychotic drugs.
All statistical analyses were performed using the Statistical Analysis System (version 8.02; SAS Institute Inc, Cary, NC). To address the first prediction, comparisons of outcome variables were performed among the 3 groups (no cannabis, bipolar first, and cannabis first). Survival analysis was used to compare the groups on recovery from bipolar disorder and recurrence of significant affective symptoms using the log-rank χ2 test. If the omnibus comparison was significant, then pairwise comparisons were performed to further evaluate differences. Times to events (eg, recovery and relapse) were evaluated using Cox proportional hazards regression models. Potential mediator variables included the variables listed in Table 2; those that demonstrated significant between-group differences were entered as covariates in Cox and other statistical models. To extend the testing of the first prediction and test the second prediction—that cannabis use disorders would be associated with mania rather than depression—analysis of covariance models were used to compare groups on the percentage of weeks during follow-up spent in recovery (symptom severity scores of 1 or 2), with full affective syndromes (scores of 5 or 6), and with subsyndromal symptoms (scores of 3 or 4). Tukey Honest Significant Difference post hoc procedures or least squares means protected t tests were used to further examine planned pairwise comparisons in these models.
To complete the examination of the hypothesis and predictions, survival analysis was also used to examine rates of symptomatic recovery and recurrence of cannabis use disorders among groups. Cox regression models were used to examine time to events. Analysis of covariance models were used to examine differences among the groups in the percentage of time with symptoms of cannabis use disorders during follow-up (symptom severity scores >3) (Table 1). Regression models evaluated associations between the percentage of weeks with various affective symptoms and syndromes and symptoms of cannabis use disorders. For these planned comparisons, significance was defined as P<.05. Other analyses were performed as indicated for completeness.
Sixty-nine (47.9%) of the 144 patients exhibited a co-occurring cannabis use disorder, including 36 (25.0%) in whom the bipolar disorder preceded (or was concurrent with) the onset of cannabis abuse or dependence (bipolar first) and 33 (22.9%) in whom the converse was true (cannabis first). The cannabis first group was more educated and had a later age at onset of bipolar disorder than the other groups (Table 2). The cannabis first group was also significantly older than the bipolar first group (Table 2). Both comorbid groups were more likely than the no cannabis group to exhibit psychosis at the index assessment and to be male (Table 2).
Despite differences in age and age at onset of bipolar disorder, the age at onset of cannabis use disorders was not significantly different among groups (F1,68 = 3.6; P = .06). In addition to cannabis use and abuse, these patients had high rates of alcohol use disorders. Both comorbid groups exhibited higher rates of alcohol abuse and dependence than did the no cannabis group. Finally, the cannabis first patients exhibited higher rates of other drug use disorders than did the other 2 groups (Table 2).
Figure 1 illustrates survival curves for (symptomatic) recovery from bipolar disorder. Consistent with the first prediction, these curves differed significantly (log-rank χ22 = 11.8; P = .003): the cannabis first group showed higher rates of and more rapid recovery than did the other 2 groups, which did not differ (log-rank χ21 = 0.1; P = .76). In the Cox regression model, group assignment was not significantly associated with time to recovery after adjusting for potential mediator variables that significantly differed among groups—namely, age, sex, education, presence of psychosis, age at bipolar disorder onset, and history of an alcohol or other drug use disorder at the index assessment (χ21 = 1.0; P = .33) (Table 2). Among these potential mediator variables, age at bipolar disorder onset (χ21 = 5.7; P = .02) and sex (χ21 = 3.7; P = .05) were significantly associated with time to recovery. Specifically, women had a longer mean (SD) time to symptomatic recovery (62  weeks) than did men (51  weeks), and earlier age at onset was associated with slower recovery (r = −0.20). Recurrences of affective episodes were common in all the groups, and these survival curves did not differ significantly among groups (Figure 2) (log-rank χ22 = 3.7; P = .16); nor were there significant group differences in times to recurrence (Cox regression: χ21 = 0.01; P = .91).
Table 3 lists the percentage of follow-up weeks spent in mood episodes and with subsyndromal mood symptoms. Group comparisons were made after adjusting for the previously noted potential mediator variables. The 3 groups did not significantly differ in total weeks of follow-up (F2,134 = 0.7; P = .48), percentage of weeks in recovery (F2,134 = 2.3; P = .11), or weeks with subsyndromal affective symptoms (F2,134 = 0.5; P = .62). However, the bipolar first group spent a significantly greater percentage of follow-up in an affective episode than did the other groups (Table 3). Consistent with the second prediction, this difference resulted from more time in manic (F2,134 = 2.8; P = .06) and mixed (F2,134 = 3.8; P = .03) episodes, particularly compared with the no cannabis group. Finally, the 2 comorbid groups exhibited significantly more rapid cycling during follow-up than did the no cannabis group (Table 3).
Figure 3 illustrates symptomatic recovery from cannabis use disorders for the 2 comorbid groups. The 2 groups exhibited significant differences in recovery in that recovery was more common in the cannabis first group than in the bipolar first group, consistent with the initial prediction (log-rank χ21 = 13.7; P<.001). Differences in time to recovery from the Cox regression model were nearly significant after adjusting for potential mediator variables (ie, age, education, age at bipolar onset, and rates of other drug use disorders at the index assessment) (χ21 = 3.1; P = .08). More than 70% of the patients using cannabis before hospitalization did not resume use for at least 8 weeks after hospital discharge. More than 60% of both comorbid groups exhibited symptomatic recurrences of cannabis use disorders, which was almost identical between the groups (data not shown) (log-rank χ21 = 0.1; P = .99). The bipolar first group spent a greater percentage of follow-up time with symptoms of a cannabis use disorder than did the cannabis first group (Table 3).
Sixteen patients developed a new cannabis use disorder during follow-up. Compared with the no cannabis group, these patients were nonsignificantly younger (mean ± SD, 18 ± 6 years; F1,90 = 2.1; P = .15) and less educated (mean ± SD, 9 ± 3 years; F1,90 = 3.7; P = .06). They had significantly higher drug use severity scores in the month before hospitalization (mean ± SD, 0.4 ± 0.9; F1,90 = 6.9; P = .01). They were otherwise clinically similar to the no cannabis group. Most of these patients (n = 10) developed the new cannabis use disorder within the first year (mean ± SD, 55 ± 46 weeks).
To extend these primary analyses, we performed correlations among the percentage of weeks with cannabis use and bipolar disorder symptoms while adjusting for the percentage of weeks with alcohol abuse. In the combined group of comorbid patients, the percentage of weeks during follow-up in a depressive episode was significantly associated with the percentage of weeks of cannabis abuse symptoms (adjusted r = 0.29; P = .02). This association arose more strongly from the bipolar first group (adjusted r = 0.44) than from the cannabis first group (adjusted r = 0.10). In addition, the percentage of time in remission was inversely associated with the percentage of weeks with cannabis abuse symptoms (adjusted r = −0.29; P = .02), owing primarily to the association in the cannabis first group (adjusted r = −0.36) rather than the bipolar first group (adjusted r = −0.06). In the cannabis first group, the percentage of weeks with cannabis abuse symptoms was significantly associated with the percentage of time with subsyndromal affective symptoms (adjusted r = 0.38; P = .03). No other statistically significant associations were observed among these measures.
The bipolar first group spent a greater percentage of weeks during follow-up with symptoms of an alcohol use disorder than did the no cannabis group. Although the rates of alcohol use disorders were similar in the 2 comorbid groups, the cannabis first group exhibited more alcohol dependence than did the bipolar first group (χ21 = 6.3; P = .01). In addition, the 3 groups significantly differed in the percentage of weeks during follow-up that patients experienced symptoms of an alcohol use disorder (Table 3). In a previous study,13 we classified these same patients into groups based on the relative ages at onset of alcohol use and bipolar disorders. These classifications were highly correlated with the cannabis co-occurrence categories (ϕ = 0.66; P<.001) (Table 2). Moreover, during follow-up there was a significant correlation between the time spent with symptoms of cannabis and alcohol use disorders (r = 0.24; P = .04).
Classes of medications prescribed during follow-up in the 3 groups are listed in Table 4. Lithium and conventional antipsychotic agents were less commonly prescribed to the no cannabis group than to the 2 comorbid groups; otherwise, different medication classes were similarly prescribed among groups.
Although numerical differences among groups were observed in the percentage of weeks during follow-up in which patients exhibited full adherence with at least 1 putative mood stabilizer (Table 3), these differences were not significant after adjusting for the potential mediator variables previously noted (F2,143 = 1.3; P = .28). Adding this measure of treatment adherence into the Cox regression models of bipolar recovery and recurrence did not alter the findings. Similarly, adding treatment adherence to analyses of recovery and recurrence from cannabis use disorders did not alter these results.
The primary hypothesis was that the sequence of onset of bipolar and cannabis use disorders would differentially affect outcome. Specifically, similar to findings with co-occurring alcohol use disorders and consistent with the first prediction, the cannabis first group demonstrated better recovery using simple survival analysis. However, when adjusted for potential mediator variables, time to recovery did not differ among the 3 patient subgroups. Differences in age at bipolar onset and sex among the groups seemed to explain the differences observed in this outcome variable. This observation suggests that cannabis abuse might represent a precipitant for a later age of bipolar onset in some patients who might otherwise have low vulnerability for the illness. Alternatively, one could posit that cannabis abuse delays the onset of bipolar disorder in some patients, although this is inconsistent with previous studies.3 Consistent with the second prediction, cannabis abuse was associated with manic rather than depressive symptoms, at least in the bipolar first group. Specifically, patients in the bipolar first group spent more time in mixed and manic episodes than did the other groups. However, when the comorbid groups were combined, the symptoms of cannabis abuse more strongly correlated with depressive than manic symptoms, inconsistent with the prediction. Combined, the comorbid groups demonstrated a higher rate of rapid cycling compared with patients without a co-occurring cannabis use disorder, although these differences were modest. Together, these findings suggest that there may be modest effects based on the relative sequence of onset of cannabis use and bipolar disorders in patients with both conditions. In general, because many of the group differences were modest and given the potential complexity of associations, additional work is needed to further delineate how these disorders interact over time.
The present analysis identified associations between cannabis use and affective symptoms in patients with bipolar disorder but could not determine causality, that is, whether increased cannabis use leads to more affective symptoms or the converse. In previous studies,3,7 we found that almost any temporal relationship occurs between cannabis use and affective symptoms in bipolar samples. Therefore, it is likely that these same complex relationships exist in the present sample. Nonetheless, the associations observed in this study provide no support to the notion that cannabis use improves affective symptoms in patients with bipolar disorder.14,15 Instead, the converse—that cannabis use worsens the course of illness—is more strongly supported, at least in the bipolar first group.
Similar to what we observed in a previous analysis of co-occurring alcohol use and bipolar disorders from this sample,13 there seems to be a significant effect of a first psychiatric hospitalization for mania on the course of cannabis abuse. Most patients exhibit a period of abstinence from cannabis abuse in the period immediately after hospitalization. However, recurrence of cannabis abuse is rapid and common, suggesting a brief window of opportunity to aggressively address cannabis use disorders in the period after hospitalization.
Although the courses of alcohol and cannabis use disorders in these patients with bipolar disorder were significantly correlated, they were not exactly so, that is, many patients exhibited only one or the other comorbidity, and the relative ages at onset were staggered and inconsistently linked. In addition, the effects of these 2 substances on the course of bipolar disorder, and the converse, are different.3,7,13 These results suggest that these substances must be addressed separately when designing treatment programs for patients with bipolar disorder. Moreover, the courses of alcohol and cannabis use disorders may interact in complex ways that are currently poorly understood.17 Additional studies are needed to examine the full range of effects of various substances of abuse in the course of bipolar illness. The present study provides a step toward this end.
As with all clinical studies, limitations must be considered when interpreting these findings. Although the overall sample size of this study is relatively large, subgroup analyses involved, necessarily, smaller samples, which may limit the statistical power of some comparisons. This limitation is most notable late in the study, when the number of evaluable patients diminished. In addition, symptom ratings, cannabis and alcohol use, and medication adherence ratings were primarily based on patient self-report, which might introduce bias. However, there is no a priori reason to expect this to affect the relative differences among groups. Moreover, the high rates of substance use disorders at the index assessment, and the high rates of use reported during follow-up, suggest that this bias was minimal. The calendar methods used to assess cannabis and alcohol use were primarily developed for 1-month assessments, so extension to 4-month intervals might introduce error.26 Again, there is no a priori reason to expect this to differ among groups, and we have used these methods in several other studies,3,4,6,7,13 the results of which support the validity of the approach. Age-at-onset estimates are ultimately limited by the data provided by the patients, and some early symptoms may have been subtle and difficult for them to identify. Similarly, determining specifically when casual cannabis use crossed the threshold to abuse is also often difficult. This limitation might have influenced group assignment, leading to imprecise assignment and potentially minimizing group differences (type II error). Finally, recruitment occurred at a single site; consequently, these results may not generalize to other patient populations.
Despite potential limitations, to our knowledge, this is the most extensive prospective study to date specifically designed to examine interactions between co-occurring cannabis use and bipolar disorders, particularly in a relatively unique first-episode sample. These results reinforce and extend previous findings for this common co-occurrence. They also suggest a window of opportunity immediately after a first psychiatric hospitalization to develop strategies to prevent recurrences and new cases of cannabis abuse. Finally, these results will encourage future studies to examine the complex interplay among bipolar, drug use, and alcohol use disorders.
Correspondence: Stephen M. Strakowski, MD, Department of Psychiatry, University of Cincinnati College of Medicine, 231 Albert Sabin Way (ML0559), PO Box 670559, Cincinnati, OH 45267-0559 (Stephen.Strakowski@uc.edu).
Submitted for Publication: November 28, 2005; accepted March 9, 2006.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant MH58170 from the National Institute of Mental Health (Dr Strakowski).