Antidepressant Augmentation Using the NMDA-Antagonist Memantine: A Randomized, Double-Blind, Placebo-Controlled Trial

Abstract

N-Methyl-D-Aspartate (NMDA) antagonists have garnered intense interest as a novel therapy for depression since the pivotal findings that ketamine infusions produce rapid, robust, and sustained improvement in depression symptoms.^{1, 2} However, ketamine’s use is limited by the requirement for intravenous administration, the transient dissociative and perceptual disturbances that frequently accompany its administration,³ and the challenges of preserving recovery of symptoms over longer than a few weeks after a single infusion.⁴ The NMDA antagonist memantine, while of lower affinity, faster-dissociating, and exhibiting other pharmacological differences than ketamine,⁵ also has fewer side effects and is orally administered. Memantine is also already approved by the US FDA for treatment of another neuropsychiatric disorder, Alzheimer’s dementia, making it an attractive candidate for investigation.

An initial trial of memantine as monotherapy for depression was terminated early when memantine treatment failed to separate from placebo based on response or remission rate.⁶ Although this monotherapy trial did not demonstrate benefit, there are animal findings suggesting that noncompetitive NMDA receptor antagonists such as memantine work synergistically in combination with antidepressants,⁷ providing a rationale for evaluating memantine specifically as an augmentation treatment. In this paper we report the results of a randomized trial testing the hypothesis that memantine would have greater efficacy than placebo in reducing symptoms of major depressive disorder when used to augment antidepressant treatment.

METHOD

We conducted an 8-week randomized, double-blind, placebo-controlled trial ({"type":"clinical-trial","attrs":{"text":"NCT00344682","term_id":"NCT00344682"}}NCT00344682) of memantine augmentation treatment of outpatients at the University of Massachusetts Medical School, Worcester, Massachusetts from July 2006 to December 2011. Participants who were incomplete or nonresponders to their current antidepressant were randomized 1:1 to add on either flexibly-dosed memantine, 5mg–20mg/day, or placebo to their current antidepressant using a block randomization design by the University of Massachusetts Medical School Investigational Pharmacy. A printed list computer-generated from the website www.randomization.com was used with allocation concealed from participants, research staff, and investigators. All participants began on 5 mg/day and the dose was increased by 5mg/day at weekly intervals as tolerated to a maximum dose of 20 mg/day. In the absence of limiting adverse effects, this dose was achieved approximately 22 days into the eight-week (56-day) trial.

The original target recruitment was 25 patients per study arm. However, the study was terminated prior to full enrollment in 2012 at funder request, resulting in a final enrollment of 31 patients. The study was approved by the University of Massachusetts School of Medicine Institutional Review Board.

RESULTS

Of 59 patients screened by phone or in person, 31 (53%) were randomized to receive memantine (n=15) or placebo (n=16). All randomized participants received study medication and had at least 2 post-baseline assessments, and 84% of the participants completed the trial (Figure 1). 100% of memantine and 50% of placebo recipients received a study medication dose of 20 mg/day during some portion of the trial. One participant receiving memantine dropped out of the protocol for side effects, while four participants receiving placebo dropped out of the protocol for side effects, worsening of depression symptoms, or movement out of the area. Although study enrollment was terminated early, randomization balanced patients effectively on all measured confounders (Table 1).

Open in a separate window

FIGURE 1

Screening, Randomization, and Disposition of Patients with Major Depressive Disorder Incompletely Responding to Antidepressant Therapy Randomly Assigned to Treatment Augmentation with Memantine or Placebo

No significant differences were found between memantine augmentation and placebo augmentation in our primary endpoint of change in MADRS score (LOCF) across all the repeated measures, (beta =0.133, p= 0.74) using LOCF methods (Table 2A), with the nonsignificant and modest score differences observed favoring placebo (Figure 2A). An alternative, “observed case” analysis of the data was also nonsignificant (beta= 0.497, p=0.212) (Figure 2B).

Open in a separate window

Open in a separate window

Open in a separate window

FIGURE 2

Changes in Depression Rating Scale Scores for Patients with Major Depressive Disorder Assigned to Memantine or Placebo Augmentation of Antidepressant Treatment (LOCF). A. Montgomery-Asberg Depression Rating Scales (LOCF, primary outcome) B. Montgomery-Asberg Depression Rating Scales (No imputation [“Observed Case”]) C. Quick Inventory of Depressive Symptoms, Self-Rated

Mean MADRS score change (LOCF) for memantine recipients at week 8 was −7.13 +/− 6.61 (standard deviation); mean MADRS change for placebo recipients was −7.25 +/−11.14 (Table 2B). A Cohen’s d effect size for change in MADRS score from baseline through week 8 of 0.19 (favoring placebo) was observed. A larger, although nonsignificant, change in MADRS score from baseline through study completion favoring placebo was observed for the nonimputed, observed case mixed model (7.13+/−6.11 [memantine] versus 10.75+/− 10.46 [placebo]), resulting in a medium-sized effect size (d=0.69) favoring placebo. Both these effect size estimates are larger than what simple comparisons of the change scores between memantine and placebo might suggest, since score changes for each treatment group being compared to that group’s baseline standard deviation¹⁹ (which is a narrow +/− 6.92 points for placebo recipients). A comparison of effect sizes based on final MADRS scores for the memantine and placebo groups would produce much more modest effect sizes favoring placebo (d=0.02 [LOCF] and d=0.25 [observed case]). A post hoc secondary analysis comparing baseline to week 8 MADRS change for patients with the greatest baseline depression severity also did not reveal differences favoring memantine (Table 2B, Footnote e).

Similarly, no statistically significant differences were observed for the other secondary depression outcome (QIDS-SR) in repeated measure analysis (beta=−0.246, p=0.22) or at week 8 comparison (accompanied by a neglible effect size of −0.04). However, the QIDS-SR results do directionally favor memantine, and the difference in score change at week 8 shows a smaller, although still nonsignificant, p value than the MADRS outcomes (p=0.14). Nevertheless, the changes in depression rating scales over time (Figure 2A–C) suggest little evidence of efficacy for memantine compared to placebo as measured by either QIDS-SR or MADRS.

The secondary endpoints of response and remission based on MADRS score also yielded non-significant results: 13.3% (n=2) of participants receiving memantine had a response or remission by 8 weeks, versus 18.8% (n=3) of participants receiving placebo (Fisher’s exact test, p=1.00). These results produce a nonsignificant number-needed-to-treat favoring placebo of approximately 18.

No significant differences were observed using the Ham-A anxiety scale, the suicidal or delusional thinking scales, or in any reported adverse effects (Table 3). Two serious adverse effects were reported that required hospitalization, both involving worsening of pre-existing respiratory problems judged probably unrelated to study medication (one participant received memantine and one participant received placebo).

The integrity of blinding was assessed through “guess tests” administered to participants, research coordinators, and study psychiatrists. No group was able to accurately distinguish memantine from placebo (p>0.05 for all results), with the highest proportion of correct guesses only 57.1% (observed among participants receiving memantine).

DISCUSSION

In this placebo-controlled, flexibly-dosed randomized trial in which all participants assigned memantine received the highest approved dosage (20 mg/day), no statistically significant differences were observed between memantine and placebo as augmentation agents in incomplete or nonresponders to conventional antidepressants. The primary outcome, repeatedly-measured change in MADRS score, was both statistically nonsignificant (p=0.74) (although limited sample size restricts power), and a minimal effect size was observed at week 8 (Cohen’s d=0.19), favoring placebo. The MADRS observed case repeated measures analysis was also nonsignificant (p= 0.21), with a larger week 8 effect size favoring placebo. Response and remission rates were also nonsignificant and slightly favored placebo.

In contrast, change in self-rated depression (QID-SR) were also nonsignificant but favored memantine in both repeated measures analysis and at week 8. However, interpretation of the change in QIDS-SR score likely is complicated by the fact that this outcome had the greatest (albeit nonsignificant) baseline imbalance of any rating scale (3.4 points). This difference may have facilitated the observation of greater changes in QID-SR in the memantine group. The memantine groups’ baseline scores were also more variable (standard deviation of 7.01 versus 4.16), resulting in an effect size observed at week 8 that was negligible (d=−0.04) despite the more sizable changes in QIDS-SR score observed for the memantine group.

Given the early termination of our trial, we derived post hoc power estimates for our final sample size, taking into account our repeated measures design (which boosts power).²⁰ We estimate that our trial likely had the ability to discriminate an approximate effect size for the primary outcome in the range of Cohen’s d = 0.53. This is less power than would be desired. However, while both this study and the previous placebo-controlled trial of memantine monotherapy in major depressive disorder⁶ were limited by almost identically small sample size, it may be notable that both trials observed multiple outcomes that numerically favored placebo.

With a sample size falling between a typical Phase I and Phase II trial, our investigation might be viewed as consistent with the recently-proposed “quick win, fast fail” approach to drug evaluation.²¹ This approach emphasizes the value of small studies for boosting efficiency in drug development through a focus on detecting substantial effect sizes, not statistical significance, although Type 2 errors (i.e., failure to identify valuable interventions) are possible.²¹ For this trial, the observations that the observed drug versus placebo difference was minimal, few participants dropped out, and maximum titration of possible dose was achieved for all memantine-receiving participants would all likely lessen the chances of Type 2 error.

Limitations of our trial design beyond sample size included the fact that participants could be receiving any one of a substantial number of existing antidepressants, although specific antidepressants appeared to be generally balanced between treatment groups (Table 1). We also did not limit the amount of response participants may have already had to their antidepressant (as long as baseline HAM-D 17 was ≥16), nor imposed limits on the length of time participants could have received their current antidepressant (beyond a minimum of 25 days). This 25-day period was intended to replicate the earliest point (approximately 30 days) at which at patient-provider discussions might occur about next steps in pharmacotherapy for patients experiencing an incomplete or nonresponse to antidepressants. A longer preceding treatment requirement could have yielded lower placebo response rates and reduced the chance some participants were still experiencing change in depression produced by their concomitant antidepressant; however, these concerns are considerably mitigated by the observation that, for the 77% of participants for whom information on duration of current antidepressant treatment was available, only 3 of 24 (12.5%) had received antidepressants for 3 months or less prior to trial enrollment (16.7% of memantine and 8.3% of placebo recipients [Table 1]).

Like most,^{6, 22, 23} but not all,²⁴ prior trials of memantine for depression, our dosing did not exceed the maximum dose approved for human use (20 mg/day). A small pilot study (n=8),²⁴ in which 37.5% of the participants received doses of memantine of 30–40 mg/day, reported a much higher rate of treatment responders (62.5%) than either our trial (13.3%) or the placebo-controlled monotherapy trial.⁶ However, this 8-subject pilot study’s open-label design and requirement of prior positive response to antidepressant treatment may have considerably boosted responses independent of any effects from higher dosage. Furthermore, the decision whether to evaluate higher-than-currently-approved dosages of memantine for major depressive disorder may need to consider emerging toxicological literature reporting NMDA antagonist-associated neurotoxicity in some animal models of the developing brain. This literature which has prompted FDA hearings to discuss the relevance to human anesthetic use of NMDA antagonists, especially in children.^25–27 Two major reservations concerning the relevance of these animal safety studies have been raised, the first involving the need to employ excessive doses compared to human use, and the second concerning the use of “unrealistically long” (i.e., multi-day) exposures.²⁵ While the first reservation concerning dosing would still apply to potential psychiatric use (and it may be additionally reassuring that NMDA-antagonists are being investigated in adults, not children), the second reservation (concerning duration of exposure) potentially would not apply to likely psychiatric use of oral NMDA-antagonists.

Strengths of our study include the generally good balance achieved on measured participant characteristics despite small sample size, the generally low dropout rate, the success of blinding, the success in titrating all memantine recipients to maximum dose, and the use of a repeated measures design.

To our knowledge, this trial is the only randomized placebo-controlled trial of memantine as an augmentation agent added to antidepressant treatment. Two other placebo-controlled trials have reported a lack of efficacy for memantine as an augmentation agent in other psychiatric conditions (bipolar depression²² and schizophrenia²⁸).. A placebo-controlled trial of memantine as a prophylactic monotherapy for depressive symptoms among elderly patients requiring physical rehabilitation²⁹ also failed to demonstrate efficacy.

In conclusion, this randomized trial did not detect statistically significant differences between augmentation of conventional antidepressants with memantine and placebo and also did not detect substantial effect sizes favoring memantine.. While this study’s small sample size is a limitation, its findings suggest that memantine lacks substantial efficacy as an augmentation treatment for major depressive disorder.

Acknowledgments

Footnotes

REFERENCES