Does Lithium Treatment Still Work?: Title and subTitle BreakEvidence of Stable Responses Over Three Decades

Recent reports have been critical of evidence of the effectiveness of lithium or have noted poor long-term outcomes in some clinical settings, particularly since its acceptance into American psychiatry.^{1 - 9} Emerging questions include whether lithium may have been overvalued in the past in the absence of plausible alternatives, whether its limitations are being increasingly accurately perceived, and whether long-term lithium treatment has lost effectiveness since the 1970s. Several factors may contribute to a recent tendency to question the clinical value of long-term lithium treatment of bipolar manic-depressive disorders and its place in contemporary psychiatric therapeutics. Losses in effectiveness, if they occur, might arise in patients with more atypical, treatment-unresponsive illness; in patients who are undermotivated, noncompliant, disadvantaged, or otherwise difficult to work with; or in patients who represent a broader and more heterogeneous range of persons considered candidates for trials of mood-stabilizing treatment.¹⁰ Effects of case selection and diagnostic changes were similarly cited as contributing to evidence of recent losses in average long-term outcomes in schizophrenia after significant gains in the 1960s and 1970s.^{11 - 12} In addition, the perception of lithium as a potentially dangerous substance or as a socially stigmatizing treatment may also limit its acceptance by many patients and physicians and encourage the search for better alternatives.^{13 - 14}

To test the hypothesis that the clinical effectiveness of lithium in bipolar disorders has diminished systematically since the 1970s, we reviewed available research literature on lithium treatment from 1970 to the present, as well as analyzing the experience of a large clinical sample of patients with bipolar disorder entering long-term lithium treatment over the past 3 decades in a research center.

We found 11 controlled^{15 - 25} and 13 open^{26 - 38} long-term lithium treatment trials (1970-1996) for bipolar or mixed major affective disorders (cohorts of patients identified as having only unipolar recurrent major depression were excluded) that would permit estimates of recurrence rates (percentage per month) with and without lithium treatment. These studies have been evaluated elsewhere in more detail for other purposes.³⁹ Recurrence rates were compared for years before and after the introduction of DSM-III (1981) and by regression vs years reported.

To further examine possible secular changes in responses to lithium treatment in a stable environment over 3 decades, we evaluated patients at the Lucio Bini Mood Disorders Center in Cagliari, Sardinia, using diagnostic and assessment methods detailed elsewhere.^{38 - 41} This clinic has treated patients with major mood disorders since the 1970s, following the Research Diagnostic Criteria for a Selected Group of Functional Disorders and the Diagnostic and Statistical Manual of Mental Disorders, with research-quality records and regularly updated life charts of illness course; diagnoses and definitions for recurrences of mania or bipolar depression were recently updated to meet DSM-IV criteria. Data analyzed by year of starting treatment included episodes per year and percentage of time ill in mania or depression, the proportion of patients with major improvement (≥50% reduction in the percentage of time ill compared with years from illness onset to the start of sustained lithium treatment), and proportion of cases with no new episodes during treatment.

The data obtained from the 24 published reports we found were pooled,^{15 - 38} as there was no significant difference in rates obtained in controlled and open trials (F_1,22 =1.27; P = .35). All values are presented as mean ± SD unless otherwise indicated. Recurrences per month for various types of patients with manic depression who were not receiving lithium treatment remained similar over the years of study (r = 0.016; P = .63), averaging 18.7-fold higher than for those receiving treatment (percentage per month: 28.1% ± 35.9% vs 1.5% ± 0.3%; paired t = 3.40; P = .003), suggesting no clear increase of illness severity over time. Monthly recurrence rates during treatment also did not rise over the years. Instead, they averaged 5.04 times higher in older, pre–DSM-III trials (reported in 1970-1981) (percentage per month: 2.7% ± 3.4%) than in more recent reports (conducted 1982-1996) (percentage per month: 0.5% ± 0.7%; F_1,22 = 4.78; P = .04). Regression (Figure 1) also indicated moderate and inconsistent but significant decreases in monthly recurrence rates during treatment over the years (r = −0.445; P = .03), even when a small early study¹⁸ with an aberrantly high value was excluded (r = −0.474; P = .02). These results do not indicate loss of effectiveness of long-term lithium maintenance treatment but suggest some decrease in recurrence rates in recent years.

From the Sardinian clinic population, 360 patients with bipolar I (n = 220) and bipolar II (n = 140) disorders (64.7% were women) included in the study had been started on lithium maintenance treatment in the years since 1970 (essentially as monotherapy), entering at an average age of 37.4 ± 14.2 years and continuing for a minimum of 12 months (4.6 ± 4.0 years) at mean ± SD serum concentrations of 0.62 ± 0.14 mmol/L, which remained stable over the years (r = −0.080; P = .60). There was no tendency toward greater prelithium morbidity over the years (r = 0.090 and r = 0.022 for episodes per year and percentage of time ill, respectively; P≥.50 for both). Recurrences of mania or depression averaged 0.81 ± 1.12 episodes per year in patients given lithium vs 1.83 ± 2.14 episodes per year before treatment (paired t = 8.62; P<.001). The percentage of time ill averaged 17.8% ± 21.8% during treatment vs 45.7% ± 30.7% before treatment (paired t = 14.65; P<.001). Neither the frequency of new episodes (r = 0.047; P = .70) nor the percentage of time ill (r = 0.030; P = .80) during treatment varied significantly by year of entering treatment. There was also no loss of improvement in episode frequency or percentage of time ill (for both: r = 0.023 for individuals; P = .90), as illustrated across half-decades (Figure 2). We also considered early morbidity during treatment, and again found no significant change since the 1970s with respect to episodes per year or percentage of time ill in months 1 through 12 or 13 through 24 for patients given lithium (for all 4: r≤0.046; P = .70).

Some residual morbidity during lithium maintenance treatment was expected, and only 32.5% of the Sardinian patients with bipolar disorder had no new episodes during treatment. The annual proportion of such highly responsive patients did not diminish between the 1970s and 1990s (χ²₂ = 4.03;P = .20). In addition, 65.6% of all subjects showed major benefits of treatment, as indicated by a 50% or greater reduction in the percentage of time ill, and this proportion, computed annually, did not change significantly over the decades studied (χ²₂ = 0.084;P = .60) (Table 1).

Since complex, treatment-resistant cases may present in clinic populations,^{4 ,8 - 10 ,14 ,42} we considered separately Sardinian patients with psychotic features or predominately mixed states (in ≥50% of episodes prior to lithium treatment), those meeting DSM-IV criteria for a substance use disorder at any time, and those requiring occasional, brief supplementary antipsychotic medication (at least once for ≤90 days at a time). Patients with such complex forms of illness (13.3% of cases) were less lithium-responsive than others (eg, during treatment, patients with complex forms of bipolar disorder were ill 25.0% ± 25.1% of the time vs 16.7% ± 21.1% for other patients)(F_1,358 = 6.15; P = .01). Patients with such complex illnesses accumulated 2.3-fold over the years, representing 8% (14/175) of patients from 1971 through 1984 vs 18.4% (34/185) from 1985 through 1996 (χ²₁ = 8.38; P = .004).

Finally, rapid-cycling cases (≥4 episodes in any year) represented 15.6% (56/360) of subjects, with significant decreases in prevalence since the 1970s, from 19.3% to 7.0% in the late 1990s (r_s = −0.920; P = .03), probably because of more conservative use of antidepressants. Patients with rapid cycling had more illness during as well as before lithium treatment (eg, recurrence rates averaged 3.94 ± 3.70 [rapid cycling] vs 1.47 ± 1.46 [non–rapid cycling episodes per year before lithium treatment, and 1.50 ± 1.98 [rapid cycling] vs 0.690 ± 0.846 [non–rapid cycling] episodes per year during treatment [F_1,358≤24.86; for both, P<.001]). However, the annual proportion of rapid-cycling cases diminished over the years (22.3% in 1970-1984 vs 7.6% in 1985-1997; χ²₁ = 15.52; P<.001).

The present findings support the following impressions: (1) In studies reported from the 1970s through the 1990s, recurrence rates in patients not being treated have not risen. (2) Average morbidity before or during lithium treatment did not increase in a large sample of patients with DSM-IV bipolar illness in a stable clinical setting during the same era. (3) The proportion of patients with complex bipolar disorder with psychotic features, a majority of mixed episodes, or a need for supplemental antipsychotic treatment increased somewhat, whereas the proportion of patients with rapid cycling diminished. (4) Both patients with complex bipolar disorder and those with rapid-cycling bipolar disorder were less treatment responsive. (5) Overall, there was no evidence of a net loss of benefit of lithium treatment over the past 25 years. In at least in one study center, the findings support the consensus that most patients with bipolar disorder have some residual illness during lithium maintenance, but fail to indicate that manic-depressive morbidity has consistently increased over the past 3 decades, or that the proportion of those without illness while receiving lithium maintenance therapy diminished.

Evidently, the numbers of complex cases in the Sardinian sample, although rising, were limited and were counteracted by the opposite secular trend in the proportion of rapid-cycling cases. It seems that there was no net degradation of treatment responses over the years. The decrease in rapid-cycling cases over time may reflect more conservative local use of antidepressants in patients with bipolar disorder, which would limit the risks of switching and increased cycling rates.^{43 - 46} However, in other sites, a greater accumulation of patients with complex bipolar disorder or those with rapid cycling may well contribute to losses in the clinical effectiveness of available treatments.

Caveats about the present findings may complicate interpretation of apparent secular trends in treatment response. Notably, site variance may produce potentially critical differences in the diagnostic case mix, in the proportion of patients with rapid cycling or with atypical illness, in relapse criteria, and in follow-up procedures, confounding comparisons of studies across settings as well as across years.³⁹ Bias may also have been introduced in the Sardinian clinic sample by the effort to evaluate the effects of lithium treatment, such as by requiring patients to remain in treatment for 12 months or longer. In turn, long-term adherence to lithium treatment may have been facilitated by routine use of moderate serum lithium concentrations (0.60-0.75 mmol/L), reported to provide a nearly optimal risk-benefit ratio.⁴⁷ Clinical samples yielding poor adherence and responsiveness to treatment for manic-depressive disorders may reflect a differential accumulation, over time, of patients with increasingly complex and difficult forms of illness, particularly in urban referral centers.⁴² Such trends may well limit responsiveness to other treatments as well as lithium. Finally, differences between results of controlled trials and clinical practice are commonly encountered with many treatments.¹³

We suggest that the growing American urge to abandon lithium maintenance therapy as ineffective, excessively toxic, or complicated is unwarranted. No other proposed mood-stabilizing treatment has such substantial research evidence of long-term efficacy in both type I and type II bipolar disorders, as well as yielding a substantial reduction of mortality risk.^{9 ,13 - 14 ,40 - 41 ,48 - 50}

Accepted for publication October 25, 1998.

This study was supported in part by grant MH-47370 from the National Institutes of Health, Bethesda, Md (Dr Baldessarini); by awards from the National Alliance for Research on Schizophrenia and Depression, Great Neck, NY (Dr Tondo); by a European Research Center Award from the Stanley Foundation, Muscatine, Iowa (Dr Tondo); by the Bruce J. Anderson Foundation, Boston, Mass (Dr Baldessarini); and by the McLean Private Donors Neuropharmacology Research Fund, Belmont, Mass (Dr Baldessarini).

Corresponding author: Ross J. Baldessarini, MD, Mailman Research Center 312, McLean Hospital, 115 Mill St, Belmont, MA 02478 (e-mail: rjb@mclean.org).