Hispanics constitute the largest ethnic minority in the United States, comprising 14% of the nation's total population, or approximately 42.7 million people. They are also the fastest-growing group, with a 3% increase in population from July 2004 to July 2005 (1). Forty percent of Hispanics are born outside of the United States (2), and 32% of U.S. 2000 census respondents 18 years or older who spoke Spanish at home spoke English "not well" or "not at all" (3). Lower English proficiency among Hispanics may be associated with socioeconomic disadvantage, lower acculturation levels, or both (4,5). However, there are conflicting data regarding whether language proficiency or preference is associated with physical and mental health disparities (6,7,8,9).

Worldwide, major depressive disorder is the most prevalent psychiatric disorder, is the fourth leading cause of disease burden, and is responsible for the largest amount of nonfatal burden, accounting for almost 12% of all total years lived with disability (10). Several epidemiological studies have found that when sociodemographic characteristics such as sex, age, education, socioeconomic status, and acculturation levels are controlled for, no significant differences remain in the prevalence of depression among Hispanics compared with whites (11,12,13,14,15,16). Other studies have found small differences in prevalence that are ascribed to particular age or gender groups (17,18) or subgroups by country of origin (19).

Numerous studies have documented disparities in diagnosing and treating depression in various clinical settings and patient groups comprising young (20), adult (21,22,23,24,25,26,27,28,29,30,31), and older adult Hispanics (32), particularly those with lower levels of acculturation to the United States (33). Other studies have reported that Hispanics are more likely than non-Hispanic whites to receive substandard care and to have poorer outcomes from antidepressant treatment (such as fewer follow-up visits and earlier discontinuation [34]) and may prefer psychotherapy, or a combination of psychotherapy and pharmacotherapy, to pharmacotherapy alone (35,36,37,38). Even though several studies of both Spanish- and English-speaking Hispanics have been conducted, very few of them focused on the issue of language per se, although some looked at acculturation and therefore only indirectly at language. Reviewers of the literature have concluded that lower rates of diagnosis of depression among adult Hispanic Americans, as well as the poorer outcomes found for this group, may be related to language differences, disparities in health literacy, differences in cultural idioms used to indicate or describe distress, and somatic presentations (38). Although comprehensive treatment can effectively manage major depression and diminish its associated disability (39,40), clinical trials focusing on Hispanics' use of antidepressants have been rare, mostly uncontrolled, have used nonrepresentative or small samples (41,42,43,44,45), and have not systematically compared Spanish-speaking patients with English-speaking patients.

The key question this study seeks to answer is, Do Hispanic patients with major depressive disorder living in the United States who prefer to communicate in Spanish have different disease characteristics and response patterns to antidepressant treatment (as opposed to rates of depression) than Hispanic patients who prefer to communicate in English? To our knowledge, this question has not been systematically addressed, but considering the U.S. demographic characteristics and the documented health disparities for Hispanics, it is an important question to investigate. This report is a secondary analysis from the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, a large clinical trial for depression conducted in public and private psychiatric and primary care settings (46,47). STAR*D included the evaluation of Hispanic patients whose language preference was English or Spanish, and it was the largest systematic treatment trial of major depressive disorder to be completed with "real-world" patients, or those seeking treatment with very few exclusion criteria. Therefore, it provided a unique opportunity to answer the question posed above.

Study overview and organization

STAR*D aimed to define the best treatment steps for outpatients with nonpsychotic major depressive disorder whose initial treatment with the selective serotonin reuptake inhibitor citalopram did not produce an acceptable outcome (that is, remission). The rationale, methods, design, and outcome results from the first two levels of the STAR*D trial have been detailed elsewhere (46,47,48,49).

Two of the 14 regional STAR*D centers where the study was conducted had bilingual or bicultural staff members who enrolled 195 English- or Spanish-speaking Hispanic participants. Data are presented from only these two Southern California sites (representing 60% of the total Hispanic participants).

Participants and measures

From July 2001 through April 2004, STAR*D enrolled treatment-seeking outpatients 18 to 75 years of age who had a diagnosis of nonpsychotic major depressive disorder (50) and a baseline score of 14 or higher on the 17-item Hamilton Rating Scale for Depression (HRSD-17) (51). Broad inclusion and minimal exclusion criteria were used to ensure a representative sample (46,47). All risks, benefits, and adverse events associated with STAR*D participation were explained in either English or Spanish, based on participant preference, and written informed consent was obtained before study entry. The study was approved and monitored by institutional review boards at each clinical site and by the Data Safety and Monitoring Board of the National Institute of Mental Health.

Participants identified their ethnic and racial background by using forced choices of white (Caucasian), American Indian-Alaskan Native, black (African American), Native Hawaiian-Pacific Islander, or Asian. A separate question established Hispanic or non-Hispanic identification. This study included the participants defined as being in the STAR*D level 1 analyzable sample (N=2,876) who identified themselves as white Hispanics and who sought care at clinics that had the research capability to enroll both Spanish- and English-speaking participants (47).

Participants chose whether to have their evaluations (both verbal contacts and written questionnaires) completed in English or Spanish. Bilingual and bicultural clinical research coordinators administered the instruments to those choosing Spanish language and to some of the English-speaking participants; bilingual and bicultural physicians were involved in the treatment of Spanish-speaking participants. For the purposes of this study, Spanish-speaking participants were defined as those who chose to complete all forms and interviews in Spanish.

At baseline, the following were completed: the HRSD-17, the 16-item Quick Inventory of Depressive Symptomatology-Clinician-rated (QIDS-C16) and the self-report version of the QIDS (QIDS-SR16) (52,53), as well as the 14-item Cumulative Illness Rating Scale (CIRS) (54,55) to gauge the severity and morbidity of general medical conditions. The CIRS was used to generate three scores for each participant: number of categories endorsed, total score (the sum of all categories), and severity index (total score divided by number of categories endorsed). The self-report Psychiatric Diagnostic Screening Questionnaire (PDSQ) (56,57) was also completed at baseline to determine the presence or absence of 11 potential concurrent axis I psychiatric disorders (based on a 90% specificity threshold).

Information on the primary research outcome (HRSD-17) and responses to the 30-item Inventory of Depressive Symptomatology (IDS-C30) (53) were collected by telephone-based structured interviews in English or Spanish within 72 hours of each clinic visit. Responses to items on the baseline HRSD-17 or the IDS-C30 were used to determine the presence of atypical (58), anxious (59), and melancholic (60) symptom features.

At baseline, an automated, telephone-based, interactive voice response system (61) was used to administer the perceived physical functioning and mental health functioning components of the 12-item Short-Form Health Survey (SF-12) (62), the Quality of Life Enjoyment and Satisfaction Questionnaire (63), and the Work and Social Adjustment Scale (64).

Intervention protocol

The protocol aimed to have participants reach a fully adequate dosage of citalopram for a sufficient time to ensure that those who did not reach remission (QIDS-C16 score of 5 or less on a scale of 0 to 27, with higher scores indicating more severe depression) were truly resistant to the medication. Decisions about dosage adjustments were guided by clinical assessments and side effects. Side effects were rated with the Frequency, Intensity and Burden of Side Effects Ratings (65).

Participant visits occurred at weeks 2, 4, 6, 9, and 12 (with an optional week 14 visit if needed). Remitters and responders could enter a 12-month naturalistic follow-up, although all participants without remission were encouraged to enter the subsequent randomized trial. Participants could discontinue citalopram before 12 weeks if intolerable side effects required a medication change, if an optimal dosage increase was not possible because of side effects or participant choice, or if significant symptoms (QIDS-C16 score of 9 or higher) were present after nine weeks at a maximally tolerated dosage. Participants could opt to move to the next treatment level if there were intolerable side effects or if the QIDS-C16 score was greater than 5 after an adequate trial in terms of dosage and duration.

In STAR*D, attrition was defined as exiting from the study after at least one baseline visit and before the 12-week visit (66). Participants in the nonattrition group were defined as those who discontinued citalopram treatment at any time in order to move to the follow-up or to the next treatment level, those who left for medical reasons, and those who left after completing 12 weeks of treatment.

After exit from level 1 (citalopram treatment), participants who either achieved remission or who chose not to continue on to subsequent levels were followed for up to one year under more naturalistic conditions. They were encouraged to stay on their medication, although medication and dosage changes were allowed, as was psychotherapy. Attempts were made to complete evaluations at two-month intervals for a period of 12 months. Relapse was defined as a QIDS-SR16 score (by interactive voice response system) of 11 or higher on a scale from 0 to 27, with higher scores indicating greater depressive symptom severity.

Statistical analysis

Data are presented as percentages or mean±SD values. A chi square test was used to compare discrete variables by participants' language preference. Comparisons of continuous measures by language were made with the appropriate parametric (t test) or nonparametric (Wilcoxon) test.

Remission was defined as an exit HRSD-17 score of 7 or less on a scale from 0 to 52, with higher scores indicating more severe depression (or the last observed QIDS-SR16 score of 5 or less). Summary statistics of remission (based on the HRSD-17 and the QIDS-SR16), response (based on the QIDS-SR16), percentage change in the QIDS-SR16 scores, side effects, and serious adverse events are presented, grouped by language. The percentage change in QIDS-SR16 scores was calculated from the difference between the patient's score at the final visit and the baseline QIDS-SR16 divided by the baseline score. Logistic regression models were developed to estimate the association of language with remission and response, after the analysis controlled for the effect of baseline characteristics that were not balanced among the languages and clinical site.

The times to remission and response, by language, were estimated. The end point of interest was defined as the first observed point of remission or response according to the QIDS-SR16 score. The differences in cumulative proportions of those who did not reach remission or response were plotted with Kaplan-Meier curves, and log rank tests were used to compare the cumulative proportions of the two groups. Cox proportional hazards models were used to test for differences after the analysis controlled for the effect of baseline characteristics that were not balanced among the languages and clinical sites. Time to relapse during the naturalistic follow-up period was estimated with relapse defined as the first observed QIDS-SR16 score of 11 or higher. The differences in cumulative proportions of relapse were plotted with Kaplan-Meier curves, and log rank tests were used to compare the cumulative proportions of the two groups.

A two-sided p value of <.05 was considered statistically significant in the above analyses. No adjustments of p values for multiple comparisons were made, so results must be interpreted accordingly.

Sociodemographic and clinical characteristics

Baseline sociodemographic and clinical information about the evaluable sample for English-speaking (N=121, or 62%) and Spanish-speaking (N=74; 38%) participants is presented in Table 1. [A figure showing the flow of participants through the study is available as an online supplement to this article at ps.psychiatryonline.org.] Compared with English speakers, on average Spanish speakers were nine years older, were more likely to be women, were less educated, had lower monthly incomes, and were more likely to be widowed, divorced, or separated (Table 1). There were no differences between groups in employment status or medical insurance status. By the definition of attrition given above, 32% of Spanish speakers, compared with 21% of English speakers, left the trial within 12 weeks (p=.089).

Spanish speakers tended to have their first major depressive episode later in life than English speakers and were less likely to have their first major depressive episode before age 18 (Table 1). Spanish speakers had, on average, a somewhat longer duration of their index depressive episode (2.3 months longer), and they had fewer episodes of major depression. There was no difference between groups in the percentages of those whose current depressive episode met criteria for chronic depression, defined as lasting greater than two years, or those who had recurrences. On the SF-12, Spanish speakers rated their mental functioning higher, but their physical functioning lower, than that of English speakers. There were no differences between groups in quality of life or work satisfaction. Compared with Spanish speakers, more than three times as many English speakers were treated in psychiatric clinics.

English speakers were nearly four times as likely to have made a prior suicide attempt and had more family history of mood disorder and substance abuse. The severity of baseline depression, as measured by the HRSD-17 and the IDS-C30, was not significantly different between groups. There was a significant statistical, but not clinically meaningful, difference on the QIDS-SR16, with the Spanish speakers rating themselves as more depressed. There were no differences in rates of anxious, atypical, or melancholic subtypes.

Although Spanish speakers had fewer categories of general medical conditions, their overall severity index was somewhat higher than that of English speakers, reflecting greater medical burden (Table 2). Spanish speakers had increased rates of somatoform disorder, panic disorder, and hypochondriasis, but the total number of comorbid psychiatric illnesses was not different between groups.

Treatment features

There were no differences between groups in the amount of time to first baseline visit, time spent in treatment, number of visits made, exit dosage of citalopram (an average of 41.9 mg per day), side effect burden, or attrition because of intolerance of side effects or adverse events. There were few serious adverse events and no differences between groups (Table 2).

Treatment outcome

Before adjustment for any baseline differences, more than twice as many English speakers as Spanish speakers reached remission, according to the primary outcome measure, the HRSD-17 (English speakers, 26%; Spanish speakers, 12%). On the measure of response, the QIDS-SR16, 31% of Spanish speakers, compared with 50% of English speakers, were responders. However, after adjustment for baseline differences, including sociodemographic, clinical, functional, and severity variables (Table 3), no outcome differences remained between groups.

The times to both remission and response were shorter among English speakers than Spanish speakers (Figure 1 and Table 4). However, like remission and response rates, these effects were not significant after adjustment. There were no significant differences in time to relapse between groups.

The Hispanic participants in this study were recruited from two Southern California sites, comprising 56% and 28% of enrollees at these sites. Most were presumably of Mexican or Central American origin. Language proficiency, nativity, and level of acculturation were not assessed, and group assignment was made by language preference only. STAR*D enrolled only patients who were already seeking treatment, and it provided care and medication at no additional charge, thereby obviating some of the many barriers to care facing Spanish-speaking patients (67,68). Although more Spanish speakers than English speakers left the trial before 12 weeks, this was not a statistically significant difference. Data from the Medical Expenditure Panel Survey for 1996—2001 showed that Hispanics overall had a higher rate than whites of discontinuation of antidepressants during the first 30 days of treatment (69). Lower education and lower income (both seen in our Spanish-speaking group) were also associated with early discontinuation. In the full STAR*D study, there was also a trend for Hispanics, regardless of language preference, to leave the trial at an earlier time than whites (66). Taken together, these factors argue for increased clinician vigilance when assessing Hispanic patients' adherence to taking medications, with Spanish speakers perhaps at greater risk of nonadherence to taking medication.

As a group, Spanish-speaking participants were more socially disadvantaged, as reflected in their lower mean educational attainment (one year past elementary school compared with almost completing high school by English speakers) and lower monthly incomes (45% lower among Spanish speakers), although rates of current employment were not significantly different between groups. Spanish speakers were also more likely to be seen in primary care than in psychiatric settings. This is not surprising because limited English proficiency is a risk factor for reduced use of health care services, particularly mental health services (70,71). Spanish-speaking participants had somewhat more medical burden and poorer self-rated physical function, adding to the likelihood of their seeking care in medical settings. These findings are in accord with the 2001 California Health Interview Survey, which found that Hispanics with limited English proficiency had worse self-reported overall health compared with English-proficient Hispanics, a finding that did not appear to be affected by socioeconomic status (72). However, a recent study suggested that a given level of health may be interpreted differently by more recently immigrated Hispanics and that respondents may interpret the Spanish and English versions of self-rated health questionnaires differently (73). Ultimately, the reliability of these self-report questionnaires across Hispanic subgroups remains to be fully investigated.

Spanish-speaking participants appeared to have more comorbid psychiatric illness, particularly illnesses with somatic presentations, such as panic and somatoform disorders and hypochondriasis. Studies have found this presentation to be common among Hispanics (74,75,76,77), although there is some controversy about this finding, particularly when based on responses to self-rated scales (73). The results of a depression scale validated for Mexican Americans, Cuban Americans, and Puerto Ricans found that those who are primarily Spanish speaking are less likely than those who are non-Hispanic Anglo-Americans to differentiate between affective and somatic aspects of depressive feelings, a finding more prevalent among women than men (78). Our Spanish-speaking group comprised a larger percentage of women compared with the English-speaking group, further adding to the possible reasons why this group had more somatic presentations (79). Alternatively, the PDSQ, despite having been validated with Spanish speakers (80,81), may not be sensitive enough to idioms of distress and cultural nuances, which could lead to an overestimation of somatic illnesses.

The English-speaking group had more previous suicide attempts, had higher comorbid substance abuse, rated themselves as having worse mental function, and had more family history of mood disorder. This group likely comprises more Hispanics who were born in the United States or had lived in the country for a longer period. We cannot conclude this with certainty, however, because data on these dimensions were not collected. Our findings are consistent with a previous study that suggested overall mental disorder rates were higher among Hispanics proficient in English (82). Roberts (83) reported that compared with youths of other ethnicities, Hispanic adolescents had the highest rates of suicide attempts in the United States, presumably because of acculturation conflicts, and it is possible that the English speakers in our study who had attempted suicide had made the attempt in their youth.

Delva and colleagues (84) found an association between having English as a first language and greater substance abuse among Hispanic adolescents, particularly with self-identified Mexican-American adolescents. Data from the National Epidemiologic Survey on Alcohol and Related Conditions indicated that the prevalence of any alcohol use disorder among U.S.-born Mexican Americans was nearly two times greater than the prevalence among foreign-born Mexican Americans, and the prevalence of any drug use disorder was 8.3 times greater (85). These data are in agreement with two other large studies, the National Latino and Asian American Study and the National Comorbidity Survey Replication (86).

The finding of more family history of mood disorder among English speakers is puzzling. It is possible that stigma toward mental illness was stronger among the Spanish-speaking participants and, as a group, these participants and their families were less likely to talk about and keep track of family members with mental illness.

Despite the differences between groups, there were no meaningful differences in severity of their current depression. This is in contrast with previous studies that have reported that bilingual Latino participants expressed more severe symptoms during a Spanish-language interview (87,88), although it is unclear how many of the participants in this study were bilingual.

On the primary outcome variable and on the measure of response, English-speaking participants showed better outcomes after initial treatment, but after adjusting for the many baseline sociodemographic and clinical variables, we found no significant differences between groups. It would appear, then, that the poorer response by Spanish speakers may be related to factors such as their more disadvantaged socioeconomic status or higher medical burden, rather than their language preference per se. Vega and colleagues (71) pointed out that nearly half of the Hispanic population is born outside the United States and that very high numbers have low education and low income levels. These factors could be associated with increased interpersonal problems related to cultural and social gaps between native and foreign-born family members and increased stressors related to job instability and acculturation (89).

The extent to which lower socioeconomic status is related to poorer outcome is a subject of debate. The so-called "Hispanic paradox" implies that more recent Hispanic immigrants have lower rates of depression and other mental illness and might have better health outcomes than Hispanic immigrants who have been in the United States for longer periods, even after socioeconomic status is controlled for (90,91). Other investigators, however, provide data linking lower socioeconomic status with poorer outcomes among Hispanic immigrant groups (8,9). Although Spanish speakers were more likely to attend primary care clinics, where perhaps they would not receive optimal care, this is an unlikely explanation for the differences seen in response as previous reports from STAR*D indicate that with measurement-based care, outcomes in primary and psychiatric settings were essentially equal (47).

Another possibility could be lower levels of adherence to psychotropics within the Spanish-speaking group. Poor English proficiency is a risk factor for lower adherence to psychotropic medications (92). Even in a culturally sensitive clinic, language barriers pose difficulties in medication management with Hispanics (93). In addition, stigma surrounding the diagnosis of depression and the use of antidepressant medication may be a factor in adherence. A small qualitative study involving mostly monolingual Latinos from Mexico and the Caribbean found that perceptions of stigma emerged as one of the most common adherence issues, with antidepressant use seen as implying more severe illness, being under the effects of a drug, and weakness or failure to cope with problems (94). A report from the Medical Expenditure Panel Survey for 1996—2001, though not reporting on language preference or acculturation, indicated that antidepressant discontinuation was significantly more common among Hispanics than non-Hispanics (69). Another study, using data from a national probability survey of Hispanics in the United States, found that even after the analysis controlled for clinical and other variables, patients who reported good or excellent English proficiency were less likely to discontinue treatment (95). Unfortunately, our study made no direct measure of adherence (blood levels of citalopram or pill counts), although number of visits and final dosage were similar between groups, suggesting similar adherence to the protocol. We believe that future clinical trials of psychotropic medication should include an adequate number of Hispanic participants and measures of adherence and acculturation to clarify these points.

The preadjustment remission and response rates of the Spanish-speaking participants in this report were not only lower than those for English-speaking participants but were also lower than those for English-speaking non-Hispanic whites and blacks from the full STAR*D sample (96). It is not clear why these Spanish-speaking participants had the worst response to an initial medication, but clinicians working with Spanish speakers need to be aware of this and need to address aspects of a treatment plan over and above medication. Issues such as financial hardship, inadequate housing, economic deprivation, underemployment, poorer physical function, being away from family of origin, and acculturative stressors must all be considered in this respect.

The limitations to the overall STAR*D design, including the lack of a "usual care" or placebo group, have been discussed elsewhere (46,47,48,49) and are mitigated somewhat by the goal of determining what treatments work best after an initial treatment has proved inadequate. Specific to this study, STAR*D was not designed to specifically address the association of language and ethnicity with outcome, and the data presented herein represent a secondary analysis of data from the study. However, comparisons between ethnic groups had been planned before the study began, and therefore the methodology included enrolling large numbers of participants from Hispanic backgrounds, including both English and Spanish speakers. Another potential limitation is the amount of missing data. At the very least, the missing data reduce the power to detect difference. However, the dropout rate is similar to that reported in the full STAR*D sample and is similar to that seen in clinical trials of this type (66).

The use of self-report measures to assess comorbidities has limitations and is not as rigorous as using a structured diagnostic interview. Concerns have been raised that Spanish-language instruments might lead to misclassification of symptoms reflecting somatic versus mental disturbances (97). In addition, most, but not all, study measures in Spanish have been validated. We have no data on whether there are differences in approach to completing self-rated forms between English and Spanish speakers, but it is possible that literacy skill may influence the reliability of self-reported data.

Although participants were assessed in their language of preference, English proficiency levels were not documented. Preference for language may not be as strongly related to socioeconomic status or acculturation as language proficiency is, but we had no measure of proficiency. Some participants, particularly men, may have preferred to be interviewed in English even though they may have been more fluent in Spanish. Studies have shown that men acculturate and become fluent in English more rapidly than women, possibly as a function of participation in the labor force (98). We did not collect data on length of residence in the United States, nor did we include a measure of acculturation—something that future investigations should address. The use of language as a proxy for acculturation among Hispanics has recently been discussed in a study that looked at utilization of mental health services (4,90). No conclusions were drawn that indicate that language is the best measure of acculturation, but its utility is suggested by the significant differences in utilization found between the language groups.

We have made the assumption that pharmacological issues of drug metabolism and pharmacodynamics were similar across groups, because all participants self-identified as Hispanics. There would be no reason to assume that, solely on the basis of language preference, either group would have a different pattern of genetic polymorphisms.

By choosing to study Hispanics from only two of the 14 STAR*D regional centers, we aimed to decrease the heterogeneity of the sample. This limits the usefulness of generalizing these results to other groups, particularly to Spanish speakers from Puerto Rico, Cuba, and Caribbean countries and to those who, for whatever reason, are not treatment seekers. Also, because STAR*D required all participants to begin an antidepressant and did not have a psychotherapy arm, results may not be generalizable to Hispanics who, given their preference for nonmedication approaches, might not have enrolled in the trial. Finally, these results might not be applicable as a whole to Spanish-speaking Hispanic patients with depression, because some estimates indicate that over 40% of Spanish-speaking Hispanic citizens lack health insurance (the corresponding figure for noncitizens is over 70%) (99), a figure higher than what was seen in STAR*D.

To our knowledge, this is the first report to compare outcomes of English- and Spanish-speaking Hispanics who had major depressive disorder and were treated with an antidepressant. Spanish-speaking participants responded less well to citalopram. However, after adjustment for the baseline differences, outcomes were similar for both groups. The reasons for the preadjustment results may be related to sociodemographic, socioeconomic, or acculturative differences between groups. Nevertheless, clinicians need to be aware of the low remission rates of Spanish speakers and need to address the many comorbidities and socioeconomic difficulties that these patients face.

This project was funded under contract N01-MH-90003 from the National Institute of Mental Health (NIMH). The content of this article does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of commercial products or organizations imply endorsement by the U.S. government. The authors acknowledge the editorial support of Jon Kilner, M.S., M.A.

Dr. Lesser is supported by grants from Aspect Medical Systems, and he has spoken for Medical Education Speakers Network. Dr. Zisook has received research support from PamLab and Aspect Medical Systems and has received honoraria from GlaxoSmithKline and AstraZeneca. Dr. Wisniewski is a consultant for Bristol-Myers Squibb and Organon. Dr. Cook has received research support from Aspect Medical Systems, Cyberonics, Eli Lilly and Company, Novartis, and Sepracor. He is a consultant to Scale Venture Partners and is on the speaker's bureau for CME LLC, Medical Education Speakers Network, and Wyeth. Dr. Trivedi receives research support from or is a consultant for AstraZeneca, Bristol-Myers Squibb, Cephalon, Eli Lilly and Company, Fabre-Kramer Pharmaceuticals, Forest Pharmaceuticals, GlaxoSmithKline, Janssen Pharmaceutica, Johnson & Johnson, Neuronetics, Pfizer, Targacept, and Wyeth-Ayerst Laboratories. The other authors report no competing interests.