LITHIUM TREATMENT OF ALCOHOLISM

Lithium has actually been extensively studied as a treatment for alcoholism. The older literature contains four double-blind, placebocontrolled studies (Fawcett et al. 1984, 1987; Pond et al. 1981; Merry et al. 1976; Kline et al. 1974) in which patients were not selected for depression, but outcome was compared in depressed and nondepressed subgroups. A principal hypothesis of the studies was that lithium would have a direct effect on drinking behavior independent of its mood- stabilizing effects. Depression was assessed with cross-sectional scales rather than by clinical history and syndromal diagnosis. In three of the studies (Fawcett et al. 1987; Merry et al. 1976; Kline et al. 1974), lithium-treated patients showed a significant reduction in alcohol consumption compared to controls. In the fourth study this effect held only for the depressed subgroup (Merry et al. 1976). However, none of the studies detected a significant reduction of depression symptoms in the lithium-treated groups compared to placebo groups. Depression scores tended to be reduced at followup in all subjects, suggesting a high placebo response rate. On balance, these early findings suggested that lithium might have a salutary effect on alcoholism that might be unrelated to its mood-altering effect. Another possibility is that lithium might modulate alcohol intake via serotonergic mechanisms, reminiscent of preclinical studies suggesting that serotonin uptake inhibitors may reduce self-administration of alcohol (Gorelick 1989; Naranjo et al. 1990) and cocaine (Carroll et al. 1990; Kleven and Woolverton 1993). The suggestion is instructive because it highlights the possibility that an antidepressant medication might directly influence the pathophysiology of addiction quite apart from any effects on mood. Subsequently, a large and well-designed multicenter trial failed to detect any effect of lithium on drinking behavior, even in the subsample with syndromal major depression (Dorus et al. 1989). This failure to replicate is consistent with several possibilities, including that of a very modest efficacy of lithium. However, the findings are inconclusive on the notion of a medication-responsive depressed subgroup. The data indicate that a lithium-responsive depressed subgroup seems unlikely, but lithium is at best a modestly effective antidepressant. Antidepressant Treatment in Cocaine Abuse

In an approach reminiscent of studies on lithium in alcoholism, a number of placebo-controlled trials of desipramine (Gawin et al. 1989; Giannini and Billett 1987; Weddington et al. 1991; Arndt et al. 1992; Kosten et al. 1992; Carroll et al. 1994) have been conducted to test whether desipra-mine has a direct effect on cocaine use behavior, independent of its antidepressant effects. This hypothesis stems in part from preclinical studies showing that tricyclic antidepressants reverse stimulant-induced changes in intracranial self-stimulation (Markou et al. 1992). Depression was not an inclusion criterion for these studies. In the first large trial in this series, Gawin and colleagues (1989) showed a robust, favorable effect of desipramine on cocaine use and craving. The effect was not diminished when the small subgroup with major depression was removed from the analysis, suggesting that it was indeed independent of any antidepressant effects. Large placebocontrolled trials have subsequently failed to replicate the robust effect, although several suggested that a desipramine effect may occur early in treatment (Kosten et al. 1992, Carroll et al. 1994) or in the mildly ill subgroup (Carroll et al. 1994). Of interest to this review, a secondary analysis of one of the studies (Ziedonis and Kosten 1991) showed a favorable effect of desipramine on cocaine use in the subsample with depression at baseline. Also several studies demonstrated a desipramine effect on mood (Giannini and Billett 1987) or psychological symptoms (Arndt et al. 1992). Similarly, preliminary analysis of a study by the authors of imipramine for cocaine abuse suggests that tricyclic antidepressants may reduce cocaine use in the subsample with depression (Nunes et al. 1995). Fluoxetine also showed some promise for treatment of cocaine abuse in initial open (Batki et al. 1991) and placebo-controlled trials (Batki et al. 1993), although subsequent placebo-controlled trials (Covi et al. 1995, Grabowsky et al. 1995) have failed to replicate these findings. Once again, depression was not an inclusion criterion. To date, no study of either desipramine or fluoxetine has focused exclusively on depressed cocaine abusers. Antidepressant Treatment in Opiate Addiction

During the 1970's and early 1980's, six randomized, double-blind, placebo-controlled trials of tricyclic antidepressants in methadonemaintained or detoxifying opiate addicts were reported (Kleber et al. 1993; Woody et al. 1982, 1975; Goldstein et al. 1992; Titievsky et al. 1982; Batki et al. 1987). These studies, summarized in table 2, were stronger methodologically than the older literature on TCAs in alcoholism, and in contrast to the studies on antidepressants for cocaine abuse, they all selected subjects with evidence of current depression. All but one (Titievsky et al. 1982) measured both mood and drug use outcomes. Sample sizes were mainly small (N < 50). Trial lengths were usually 4 weeks, which is adequate, although minimally so (Quitkin et al. 1984). Doses of tricyclics were low (Quitkin 1985). However, methadone slows the metabolism of TCAs (Kosten et al. 1990), so that adequate blood levels may actually have been achieved, even though blood level monitoring was not employed. As shown in table 2, four of five studies employing doxepin report superiority to placebo on measures of depression (Goldstein et al. 1982; Woody et al. 1975; Titievsky et al. 1982; Batki et al. 1987). Three found medication superior to placebo on at least a few selfreport drug abuse outcome measures (Woody et al. 1982, 1975; Batki et al. 1987). However, another three either lacked drug abuse outcome measures (Titievsky et al. 1982) or detected no medication versus placebo differences on drug abuse measures (Kleber et al. 1983; Goldstein et al. 1982). No studies found doxepin effects on urine toxicology. Thus, the evidence for a doxepin effect on drug abuse is equivocal. In a three-group design, Woody and associates (1982) found trends suggesting that desipramine was no better than placebo and inferior to doxepin, although the sample size (N = 10 per cell) was too small for a conclusive analysis. Another study found that imipramine had no effect on either depression or drug abuse (Kleber et al. 1983). Because imipramine and desipramine are relatively nonsedating TCAs, the suggestion is that doxepin may work in this population through nonspecific sedative effects (Kleber et al. 1983; Woody et al. 1982). In summary, several of these studies demonstrated clear-cut antidepressant effects of TCAs in opiate addicts, but none demonstrated clear effects on drug abuse outcome. These findings suggest that depression can be identified and treated in methadone maintained opiate addicts, but is, at least in many cases, an independent disorder that does not contribute to the etiology of addiction. It is also possible that methodologic shortcomings, including inadequate doses and trial lengths, contributed to a failure to demonstrate decreased drug use. Further, the placebo groups improved substantially in three of the trials (Kleber et al. 1983; Woody et al. 1982; Goldstein et al. 1982). A high placebo response rate makes detection of medication effects difficult, suggesting that the samples were heterogeneous and primarily evidenced transient dysphorias that resolved spontaneously. All studies studied depressed patients using cross-sectional symptom scales rather than clinical history and syndromal diagnosis. As noted, depression in opiate addicts is often transient (Rounsaville 1986a; DeLeon et al. 1973), which suggests that antidepressant treatment of homogeneous samples with primary or chronic affective disease would yield more robust effects on mood and drug use. With the benefit of hindsight on earlier design shortcomings, the authors and colleagues have sought to test the effectiveness of a standard antidepressant, imipramine, for the treatment of depression and drug abuse in methadone patients. Designs included minimum adequate trial lengths of at least 6 weeks, dosages of 150 to 300 mg/day, and selection of depressed subjects through diagnostic interviews by experienced clinicians who applied stringent criteria for current and lifetime DSM-III-R depressive syndromes.