Plasma GABA predicts acute response to divalproex in mania

Plasma GAB A Predicts Acute Response to Divalproex in Mania Frederick Petty, A. John Rush, John M. Davis, Joseph R. Calabrese, Susan E. Kimmel, Gerald L. Kramer, Joyce G. Small, Marvin J. Miller, Alan E. Swann, Paul J. Orsulak, Molly E. Blake, and Charles L. Bowden

Bipolar L manic phase inpatients were treated with divalproex sodium, lithium, or placebo in a previously reported parallel group multicenter, double-blind, randomized, controlled acute phase treatment trial. Plasma concentrations of gamma aminobutyric acid (GABA) were measured before and after treatment. Higher pretreatment plasma GABA levels were significantly (p = .04) related to a better clinical response to divalproex (n = 19). Pretreatment plasma GABA levels did not correlate with response to either lithium (n = 13) or placebo (n = 31). Following treatment with divalproex sodium, plasma GABA levels decreased significantly (p < .05), compared to placebo. Pretreatment plasma GABA levels were not related to overall severity of manic symptoms. Plasma GABA may predict response to pharmacologic agents acting on the GABA system.

Key Words: Bipolar disorder, mania, GABA, divalproex, valproic acid, lithium, response prediction B I O L PSYCHIATRY

1996;39:278-284

Introduction More than a decade ago, Emrich (1980) proposed a "GABA hypothesis of affective disorders" based on his observation that patients with bipolar illness derived a From the Veterans Affairs Medical Center and Department of Psychiatry, University of Texas Southwestern Medical Center, Dallas, TX (FP, AJR, PJO, GLK); Illinois State Psychiatric Institute and Department of Psychiatry, University of Illinois at Chicago, Chicago, IL (JMP); Case Western Reserve University School of Medicine. Cleveland, OH (JRC, SEK); Department of Psychiatry, Indiana University School of Medicine and Lame D. Carter Memorial Hospital, Indianapolis, IN (JGS, MJM); Department of Psychiatry, University of Texas Medical School, Houston, TX (AES); Abbott Laboratories, Abbott Park, IL (MEB). Department of Psychiatry, University of Texas Health Science Center and Audie L. Murphy Memorial Veterans Hospital, San Antonio, TX (CLB). Address reprint requests to Frederick Petty, Psychiatry Service 116A, Veterans Affairs Medical Center, 4500 South Lancaster Road, Dallas, TX 75216. Received October 7, 1994; revised January 12, 1995.

© 1996 Society of Biological Psychiatry

therapeutic benefit from treatment with the anticonvulsant valproate. Emrich reasoned that the therapeutic efficacy of valproate in bipolar patients was related to valproate being a GABA agonist, and that bipolar disorder might be associated with a GABA deficit. Data accumulated in the ensuing years tend to be consistent with Emrich's theory. Plasma GABA concentrations in groups of bipolar patients are significantly lower than in normal control groups, whether the patients are manic, depressed (Petty et al 1993a), or euthymic (Berrettini et al 1982, 1986). Plasma GABA concentration in humans may be a clinically useful index of brain GABA activity (see Petty et al 1993a for review). Acute administration of valproic acid increases brain, cerebrospinal fluid, and plasma concentrations of GABA in laboratory 0006-3223/96/$15.00 SSDI 0006-3223(95)00141-3

GABA and Mania

BIOLPSYCHIATRY

279

1996;39:278-284

animals (Loscher 1993). Several clinical studies, including both open and randomized clinical trials (Calabrese and Delucchi 1990, Pope et al 1991, McElroy et al 1992, Bowden et al 1994), have demonstrated the efficacy of divalproex sodium (Depakote, a stable coordination compound of sodium valproate and valproic acid in a 1:1 molar ratio) in the treatment of bipolar disorder. Based on these observations, we have investigated the relationship between plasma levels of GABA and response to divalproex, lithium, or placebo in patients experiencing acute manic episodes. Our hypotheses were that (1) patients with low plasma GABA would demonstrate a preferential response to divalproex, (2) because plasma GABA appeared to be unaffected by the mood state (Petty 1994), plasma GABA would not change significantly with treatment, and (3) plasma GABA would not be related to the severity of manic symptoms.

Methods Blood samples for plasma GABA determinations were obtained from patients enrolled in a clinical trial described elsewhere (Bowden et al 1994). Briefly, the study was a randomized, 21-day, multicenter, double-blind, parallel group trial comparing the safety and efficacy of divalproex sodium, lithium, and placebo in the treatment of the manic phase of bipolar I disorder in hospitalized patients. All patients met Research Diagnostic Criteria (Spitzer et al 1975) for manic disorder, and all had Mania Rating Scale scores on the Schedule for Affective Disorders and Schizophrenia (SADS, Spitzer et al 1975) of at least 14, with at least four items scored -->2on the last washout day. After a 3-21 day washout period, patients were randomly assigned to receive divalproex sodium, lithium carbonate, or placebo in a 2:1:2 ratio. Divalproex was administered at an initial dose of 750 mg/day, and lithium carbonate at 900 mg/day, in divided doses three times daily, and increased every 2 to 3 days in order to achieve therapeutic serum concentrations of -----40 p~g/ml valproate or ->0.4 mEq/L lithium by Day 8 of the study. Lorazepam and/or chloral hydrate could be used within protocol-specified limits for the first 10 days of the study, but no neuroleptics were allowed. Blood samples for plasma GABA determination were drawn after an overnight fast between 0700-0900, at baseline after washout, and again at exit from study, which was Day 14-21. Plasma concentrations of GABA were determined by high pressure liquid chromatography using a modification (Petty et al 1993a) of the method of Manyan and Hare (1980). The coefficient of variation for this assay is 5% over the concentration range 50-250

pmole/ml. All laboratory determinations were independent of and blind to all clinical ratings. A total of 179 acutely manic patients were randomized into the clinical trial at eight study sites. A total of 102 patients completed the 3-week protocol. Of these, 63 patients (19 divalproex; 13 lithium; 31 placebo) at 6 sites provided both pre- and post-study blood samples for plasma GABA that were suitable for analysis. One site experienced problems with its storage freezer, leading to sample deterioration and spuriously elevated plasma GABA values in seven subjects, had plasma GABA levels over 250 pmole/ml (the highest level ever documented by our laboratory) at one or both samplings, and were not included in the data analysis. Three other patients were excluded from data analysis because their serum concentrations of valproic acid were not in the therapeutic range at exit from study. The symptom severity was assessed by the Manic Syndrome Score (MSS), which corresponds to the sum of the five SADS items for elevated mood, less need for sleep, excessive energy, excessive activity, and grandiosity (range 0-30). We tested hypothesis (1) by Pearson product moment correlation relating baseline pre-treatment plasma GABA levels to percent change (pretreatment vs. posttreatment) in the MSS and by correlation between baseline pretreatment plasma GABA levels and end of treatment raw scores on the MSS for each treatment group. We tested hypothesis (2) by comparison of the change from pretreatment to posttreatment plasma GABA levels, and also of posttreatment GABA levels, between treatment groups with the analysis of variance, using Dunnett's post hoc comparison test with placebo as the control group. Serum drug concentrations were not included in the statistical analysis, since the protocol required that all patients have their dosage increased as tolerated to a predetermined level, resulting in a narrow range of drug concentrations. We tested hypothesis (3) by Pearson product moment correlation relating pretreatment baseline plasma GABA levels to pretreatment MSS scores for the entire patient sample. All tests were two-tailed, and analyses were performed using the Systat (Systat, Inc., Evanston, IL).

Results Treatment groups were not statistically different with respect to demographic characteristics (Table 1).

Hypothesis 1 The relationship between pretreatment plasma GABA levels and clinical response was evaluated by Pearson product moment correlations. A significant negative correla-

280

BIOLPSYCHIATRY

F. Petty et al

1996;39:278-284

Table 1. Summary of Patient Demographics Treatment group

Variable

Divalproex (N = 19)

Lithium (N = 13)

Sex Female Male

14 5

(74%) (26%)

5 8

(38%) (62%)

14 17

13 6

(68%) (32%)

8 5

(62%) (38%)

21 (68%) 10 (32%)

36.5 (11.3) 19-57

38.6 (10.4) 25-59

Race Caucasian Non-Caucasian Age (years) Mean (SD) Range

37.3 (11.6) 20-58

tion (r = - . 4 7 ; p = .04) was found between baseline plasma G A B A concentrations and percent change (pre- to posttreatment) in MSS for the divalproex group, but not for the lithium (r = .30; p = .54) or placebo (r = .29; p = .11) groups (Figure 1). Similarly, baseline plasma G A B A had a significant negative correlation with posttreatment MSS for the divalproex treated patients (r = - . 5 2 ; p = .02), but not for those given lithium (r = .19; p = .55) or placebo (.21; p = .10) (data not shown). Patients with higher baseline plasma G A B A levels had a better response to divalproex than those with lower levels, but baseline plasma G A B A levels did not correlate with response to lithium or to placebo.

Hypothesis 2 Baseline plasma G A B A levels were not significantly different among the three groups [F(2,60) = 0.21, p = .81]. Change in plasma G A B A concentrations pretreatment to posttreatment was significantly different among the three groups [F(2,60) = 5.46, p = .007], and lower in the divalproex group (p = .004) but not the lithium group (p = .17) (Figure 2, Table 2), when compared to the placebo group. Posttreatment plasma G A B A concentrations were significantly different among the three groups [F(2,60) = 9.35, p < .001], and lower in both the divalproex group (p < .001) and the lithium group (p = .018), when compared to the placebo group (Table 2). Baseline MSS was not significantly different among the three groups [F(2,60) = 0.47, p = .63], nor was change in MSS [F(2,60) = 1.96, p = .15]. Posttreatment MSS was significantly different among the three groups [F(2,60) = 3.25, p = .046], and lower in the divalproex group (p = .028) but not the lithium group (p = .38), when compared to the placebo group.

Placebo (N = 31)

(45%) (55%)

Hypothesis 3 The correlation between baseline plasma G A B A levels and severity of mania, rated at baseline by the MSS for all patients, was not significant (n = 63, r = 0.12, p = 0.36).

Discussion Use of a placebo in acute phase treatment of manic patients is controversial, but necessary in testing new treatments because of the episodic nature of the disorder, leading to spontaneous improvement in a significant proportion of patients. Patient safety was protected by inpatient status and daily monitoring of each patient's clinical state, use of rescue medication, and by the rapid termination from study any patient who significantly worsened, became dangerous, or requested standard treatment. The mean plasma G A B A levels measured in the present work are somewhat higher than those we have previously reported for patients with bipolar disorder (Petty et al 1993a). Plasma G A B A levels are increased at room temperature if blood specimens are not promptly centrifuged and frozen, due to hydrolysis of homocarnosine and other G A B A containing dipeptides, and due to lysis of platelets. Insuring standard sample preparation in a multicenter study such as this is difficult, and the present findings need to be replicated, preferably in one center. With regard to the hypotheses tested, we found that: (1) Patients with higher baseline plasma G A B A levels generally had a greater clinical response to divalproex. Baseline plasma G A B A concentrations were related to clinical improvement in patients treated with divalproex, but not in patients treated with lithium or placebo. (2) Plasma G A B A concentrations decreased significantly during treatment with divalproex compared to placebo treated patients; posttreatment plasma G A B A concentrations were signifi-

GABA and Mania

BIOL PSYCHIATRY 1996;39:278-284

DIVALPROEX

281

LITHIUM

40

150

20 •

100

0

"

~o

-60

/

•

-

•

%

-5o

-80 -100

i 80

60

~ 100

A.

i 120

i 140

J 160

m 180

i 200

i 220

BaselinepGABA(pmol/ml)

-100 B.

'

I

I

60

80

100

I

I

I

I

120 140 160 180 BaselinepGABA(pmol/ml)

I,

I

200

220

PLACEBO 60 40 20

._~

"o

o

. ~ -20

Figure 1. Percent change in Manic Syndrome Scale (MSS) as a function of baseline pretreatment plasma GABA levels in patients treated with (A) divalproex, (B) lithium, and (C) placebo.

/ ~ -40 -60 -80 -100 60 C.

I

I

80

I00

I

I

I

I

I

120 140 160 180 200 BaselinepGABA(pmol/ml)

I

f

220

240

cantly lower than placebo for both divalproex and lithium treated patients. (3) Plasma G A B A did not correlate with overall severity of manic symptoms before treatment. W e had hypothesized that patients with lower plasma G A B A levels would preferentially respond to divalproex, and that plasma G A B A levels would not change significantly during treatment. Therefore, it was unexpected that high pretreatment levels of plasma G A B A would predict a better clinical response to divalproex, and that acute phase

treatment with the G A B A agonist divalproex would decrease plasma G A B A levels. As noted, low G A B A has been proposed to have an etiologic role in bipolar disorder. Therefore, it seems counterintuitive that decreasing a neurochemical which is hypothesized to be deficient could be beneficial. With regard to the decrease in plasma G A B A concentrations upon treatment with divalproex sodium, there are preliminary data on plasma G A B A and several somatic

282

BIOLPSYCHIATRY

F. Petty et al

1996;39:278-284

LITHIUM

DIVALPROEX

220

220

200

200

180

180

16o

"5 E

m< 140 " 120 100

T

E -6 160 E Q. v

i

t

80

140 ,< (.9 o. 120 100 80 6O

6O

BASELINE pGABA

A.

FINAL pGABA

B.

t I

I

BASELINE pGABA

FINAL pGABA

PLACEBO

240

220 200 .180

~ 160 ~140 120 100

t

Figure 2. Plasma GABA concentrations at baseline pretreatment and final posttreatment in patients treated with (A) divalproex, (B) lithium, and (C) placebo.

80 60

12.

I

I

BASELINE pGABA

FINAL pGABA

treatments. Electroconvulsive therapy (Devanand et al 1994), diazepam (Roy-Byrne et al 1992), and alprazolam (Wolkowitz et al 1993) decreased plasma GABA, while desipramine caused no change (Petty et al 1993b). Lithium may increase plasma GABA (Berrettini et al 1983), although in the present work, no change in plasma GABA for lithium treated patients was observed. Therefore, somatic treatments for mood disorders are generally associated with decreased or unchanged plasma GABA levels. Acute neurochemical effects of valproate on the GABA

system are complex (Loscher 1993). Although increases in brain tissue levels of GABA are well documented after administration of valproate in several brain regions, the changes are minimal at clinically relevant doses (Loscher and Horstermann 1994). GABA concentrations in synaptosomes and nerve terminal are increased significantly by valproate at low doses (Loscher and Vetter 1985). Activity of GABA-transaminase, the GABA metabolizing enzyme, is decreased with acute valproate (Loscher 1993). Valproate enhances the function of GABA A receptors at the

GABA and Mania

BIOLPSYCHIATRY 1996;39:278-284

283

Table 2. Plasma GABA Concentrations and Manic Syndrome Scores Baseline GABA Treatment group Divalproex (N = 19) Lithium (N = 13) Placebo (N = 31)

Mean (pmol/ml)

(SD)

Mean (pmo~ml)

(SD)

Mean (pmo~ml)

(SD)

138 134 141

(35.5) (35.2) (31.9)

114a 123 a 152

(26.8) (33.1) (33.6)

-25 a -11 + 10

(39.7) (33.0) (37.1)

Baseline MSS Treatment group Divalproex (N = 19) Lithium (N = 13) Placebo (N = 31)

Change from baseline

Poststudy GABA

Poststudy MSS

Change from baseline

Mean

(SD)

Mean

(SD)

Mean

(SD)

18.3 17.5 18.9

(4.0) (6.3) (3.8)

10.3 ~ 12.3 14.8

(5.7) (7.8) (5.7)

-7.9 -5.2 --4.1

(6.1) (8.3) (6.3)

aSignificantly different from placebo (p < 0.05), see text for details.

level of the chloride channel (Concas et al 1991). The effect of valproate on in vivo release of GABA in ventral hippocampus, studied with microdialysis, was biphasic, with low doses reducing and high doses increasing GABA levels in extracellular fluid (Biggs et al 1992). When perfused directly into the brain, valproate inhibited GABA release in the substantia nigra (Wolf and Tscherne 1994). This result is discrepant from most other reports that valproate enhances GABA turnover in the same region (Loscher 1989). Therefore, the previous notion that valproate was simply a specific "GABA-mimetic" drug is in the process of revision, especially since the effects of valproate on other neurotransmitter systems, including the biogenic amines, appear to be more diverse than previously thought (Loscher 1993). Of perhaps greater relevance to bipolar illness are the chronic effects of valproate administration on the GABA system, which are largely unknown. Chronic valproate upregulates the GABA a receptor in hippocampus, as do chronic lithium and carbamazepine (Motohashi 1992). In humans, subchronic (4-day) administration of valproate increased plasma GABA in healthy controls (Loscher and Schmidt 1980) and increased cerebral spinal fluid (CSF) (Loscher and Siemes 1985) and plasma (Loscher and Schmidt 1981) GABA in epileptics. However, to our knowledge, a study of the chronic effects of valproic acid on brain, CSF and plasma GABA has not been reported. We have proposed that low plasma GABA represents a traitlike marker for bipolar disorder (Petty et al 1993a), and therefore hypothesized that plasma GABA levels would not change during treatment. In the present work, manic symptoms improved while plasma GABA concentrations decreased during pharmacological treatment of mania with divalproex. This provides evidence that low plasma GABA concentrations are n o t a state marker for bipolar illness, since if low plasma GABA were a state

dependent marker for the manic state, plasma GABA would increase, not decrease, with treatment and clinical improvement. The third hypothesis, that plasma GABA would not correlate with baseline manic symptom severity before treatment, was supported. A biological marker for the manic state would be expected to correlate with severity of the clinical state of mania. This finding is consistent with the proposition that plasma GABA levels are not affected by the state of illness (manic, depressed, or euthymic). We have previously shown that plasma GABA is not significantly influenced by age (in adults), gender, menstrual cycle, diet, exercise (Petty et al 1987), time of day, or season (Petty et al 1992a). Also, we have previously shown that plasma GABA does not correlate with severity of symptoms in non-bipolar major depressive disorder (Petty et al 1992b), and does not change with clinical improvement in depression (Petty 1994). Taken together, the data support the idea that plasma GABA may represent a reasonably stable biological marker. A prospective study is needed to precisely determine the state vs. trait characteristics of plasma GABA in mood disorders. In summary, the data showed that high plasma GABA was associated with better response to divalproex in mania, that plasma GABA decreased during pharmacological treatment of mania, and that plasma GABA did not correlate with severity of manic symptoms. The precise biological mechanisms involved await further research. This study was supported by the Department of Veterans Affairs Medical Research Service, NIH Research Grants MH37899 and MH41115, Abbott Laboratories, and the John Schermerhorn Fund. We thank David Morris, Ph.D., for statistical consultation; Dinah Turner-Knight for excellent secretarial assistance; and Kenneth Z. Altshuler, M.D., Stanton Sharp Distinguished Chair and Chairman, for administrative support.

284

BIOL PSYCHIATRY 1996;39:278-284

F. Petty et al

References Berrettini WH, Nurnberger JI, Hare T, Gershon ES, Post RM (1982): Plasma and CSF GABA in affective illness. Br J Psychiatry 141:483-487. Berrettini WH, Nurnberger JI, Hare TA, Simmons-Ailing S, Gershon ES, Post RM (1983): Reduced plasma and CSF gamma-aminobutyric acid in affective illness: effect of lithium carbonate. Biol Psychiatry 18:185-195. Berrettini WH, Nurnberger JI, Hare TA, Simmons-Ailing S, Gershon ES (1986): CSF GABA in euthymic manic-depressive patients and controls. Biol Psychiatry 21:842-844. Biggs CS, Pearce BR, Fowler LJ, Whitton PS (1992): The effect of sodium valproate on extracellular GABA and other amino acids in the rat ventral hippocampus--an in vivo microdialysis study. Brain Res 594:138-142. Bowden CL, Brugger AM, Swarm AC, Calabrese JR, Janicak PG, Petty F, Dilsaver SC, Davis JM, Rush AJ, Small JG, Garzatrevino ES, Risch SC et al (1994): Efficacy of divalproex vs. lithium and placebo in the treatment of mania. JAMA 271:918-924. Calabrese JR, Delucchi GA (1990): Spectrum of efficacy of valproate in 55 patients with rapid-cycling bipolar disorder. Am J Psychiatry 147:431-434. Concas A, Mascia MP, Sanna E, Santoro G, Sarra M, Biggio G (1991): In vivo administration of valproate decreases t-[S]butylbicyclophosphorthionate binding in rat brain. Naunyn-Schmied Arch Pharmacol 343:296-300. Devanand D, Shapira B, Petty F, Kramer G, Fitzsimons L, Lerer B, Sackeim HA (1995): Effects of electroconvulsive therapy on plasma GABA. Convulsive Ther, In press. Emrich HM, Zerssen DV, Kissling W, Moiler H-J, Windorfer A (1980): Effect of sodium valproate on mania. Arch Psychiatr Nervenkr 229:1-16. Loscher W (1989): Valproate enhances GABA turnover in the substantia nigra. Brain Res 501:198-203. Loscher W (1993): Effects of the antiepileptic drug valproate on metabolism and function of inhibitory and excitatory amino acids in the brain. Neurochem Res 18:485-502. Loscher W, Horsterrnann D (1994): Differential effects of vigabatrin, gamma-acetylenic GABA, aminooxyacetic acid, and valproate on levels of various amino acids in rat brain regions and plasma. Naunyn-Schmied Arch Pharrnacol 349: 270-278. Loscher W, Schmidt D (1980): Increase of human plasma GABA by sodium valproate. Epilepsia 21(6):611-615. Loscher W, Schmidt D (1981): Plasma GABA levels in neurological patients under treatment with valproic acid. Life Sci 28:2383-2388. Loscher W, Siemes H (1985): Cerebrospinal fluid gammaaminobutyric acid levels in children with different types of

epilepsy: Effect of anticonvulsant treatment. Epilepsia 26: 314-319. Loscher W, Vetter M (1985): In vivo effects of aminooxyacetic acid and valproic acid on nerve terminal (synaptosomal) GABA levels in discrete brain areas of the rat. Biochem P harmacol 34:1747-1756. Manyam NVB, Hare TA (1980): Rapid and sensitive ionexchange fluorometfic measurement of gamma-aminobutyric acid in physiological fluids. Anal Biochem 101:349-355. McElroy SL, Keck PE, Pope HG, Hudson JI (1992): Valproate in the treatment of bipolar disorder: Literature review and clinical guidelines. J Clin Psychopharmacol 12:42S-52S. Motohashi N (1992): GABA receptor alterations after chronic lithium administration--comparison with carbamazepine and sodium valproate. Prog Neuro-Psych Biol Psych 16:571-579. Petty F, Kramer G, Feldman M (1987): Is plasma GABA of peripheral origin? Biol Psychiatry 22:725-735. Petty F, Kramer G (1992a): Stability of plasma GABA gammaAminobutyric acid with time in healthy controls. Biol Psychiatry 31:743-745. Petty F, Kramer GL, Gullion CM, Rush AJ (1992b): Low plasma GABA in male patients with depression. Biol Psychiatry 32:354-363. Petty F, Kramer GL, Fulton M, Moeller FG, Rush AJ (1993a): Low plasma GABA is a trait-like marker for bipolar illness. Neuropsychopharmacology 9:125-132. Petty F, Steinberg J, Kramer GL, Fulton M, Moeller FG (1993b): Desipramine does not alter plasma GABA in patients with major depression. J Affective Dis 29:53-56. Petty F (1994): Plasma GABA and mood disorders: A blood test for manic depressive disease? Clin Chem 40:296-302. Pope HG Jr, McElroy SL, Keck PE, Hudson JI (1991): Valproate treatment in acute mania, a placebo-controlled study. Arch Gen Psychiatry 48:62-68. Roy-Byrne PP, Cowley DS, Hommer D, Greenblatt DJ, Kramer GL, Petty F (1992): Effect of acute and chronic benzodiazepines on plasma GABA in anxious patients and controls. Psychopharmacology 109:153-156. Spitzer RL, Endicott J, Robbins E (1975): Research Diagnostic Criteria (RDC) for a Selected Group of Functional Disorders, 2nd ed. New York, NY: New York Psychiatric Institute. Wolf R, Tscherne U (1994): Valproate effect on gammaaminobutyric acid release in pars reticulata of substantia nigra--combination of push-pull perfusion and fluorescence histochemistry. Epilepsia 35:226-233. Wolkowitz OM, Harris D, Turetsky NG, Reus VI, Johnson R, Gustafson M, Espinoza S, Petty F, Cooper TB, Sheline Y, Hargreaves WA: Alprazolam-neuroleptic treatment of schizophrenia. American Psychiatric Association Annual Meeting, San Francisco, CA, May 24-27, 1993.