OUP user menu

Efficacy and safety of triiodothyronine supplementation in patients with major depressive disorder treated with specific serotonin reuptake inhibitors

Rena Cooper-Kazaz, Bernard Lerer
DOI: http://dx.doi.org/10.1017/S1461145707008206 685-699 First published online: 1 August 2008


The thyroid hormone, triiodothyronine (T3), is used as a supplement to antidepressant treatment of major depression, to accelerate and enhance response and as an augmenter in patients who have not responded. While there is support from controlled trials and meta-analyses for the use of T3 in conjunction with tricyclic antidepressants, the evidence base for supplementation of specific serotonin reuptake inhibitors (SSRIs) with T3 is more limited. We reviewed the available literature on T3 supplementation of SSRIs including open-label studies and randomized controlled trials (RCTs). Five RCTs were identified. Three were enhancement studies in which T3 was administered concurrently with the antidepressant from the start of treatment and two were augmentation studies in which T3 was added to the antidepressant treatment of patients who had not responded. Three open augmentation studies were identified. The RCTs were too disparate in methodology to allow a meta-analysis to be performed. The enhancement studies are inconclusive in that one showed strongly positive effects of T3, one showed no effect and one showed a trend. The open augmentation studies supported an effect of T3 in SSRI non-responsive patients with some support from a large RCT; a smaller, underpowered RCT did not show efficacy. T3 was well tolerated in most of the studies and adverse effects do not seem to be an impediment to clinical use. Some of the studies identified clinical and thyroid function correlates of response that require further investigation. Further research is needed before it can be definitively established whether T3 is an effective supplement to SSRIs in patients with MDD. The appropriate timing of T3 supplementation needs to be explored and also the dose and length of treatment.

Key words
  • Acceleration
  • augmentation
  • enhancement
  • major depressive disorder
  • resistant depression
  • specific serotonin reuptake inhibitors
  • triiodothyronine


Major depressive disorder (MDD) is a severe, highly prevalent illness that has a substantial impact on public health worldwide. It is the most common and disabling affective disorder with a lifetime prevalence of 15% in the general population and twice that for women (Ustun et al., 2004). The pharmacological treatment of MDD, which is primarily based on antidepressant monotherapy, is problematic because of a universal delay in the onset of antidepressant effects (Bauer et al., 2002a; Blier and de Montigny, 1994) and the fact that achievement of maximal efficacy may require 12 wk or more (APA, 2000; Trivedi et al., 2006). This has led to a search for methods to accelerate antidepressant action. Furthermore, no more than half of patients achieve response, typically defined as a 50% improvement in symptom severity, with the first treatment they are administered (Trivedi et al., 2006) and less than a third achieve remission (Rush et al., 2006a; Trivedi et al., 2006). Patients who achieve response but remain short of remission suffer from significant residual symptoms, impairment of function and poor quality of life and are at increased risk for relapse and recurrence compared to patients who have remitted (Nierenberg and DeCecco, 2001; Rush et al., 2006b). The development of more effective antidepressant interventions that achieve higher rates of response and remission is an important priority in psychopharmacology.

The options available to optimize antidepressant treatment include replacement of the current antidepressant with a different agent (switch), addition of a potentiating agent to the ongoing therapy (augmentation), concurrent administration of two different antidepressant medications (combination) or administration of electroconvulsive therapy (ECT). The thyroid hormones, thyroxine (T4) and triiodothyronine (T3), have long been used to potentiate antidepressant treatment. Research with T3 has been more extensive and includes administration as a monotherapy (Feldmesser-Reiss, 1958; Flach et al., 1958) but more usually as a supplement to standard antidepressants. Studies with T3 have been of three types: (1) In augmentation studies, T3 is added to the therapeutic regimen of patients who have not responded to several weeks of treatment with a standard antidepressant. (2) In acceleration paradigms T3 is administered from the outset of treatment or very early in the course in conjunction with an antidepressant to achieve more rapid onset of antidepressant effects. (3) In enhancement studies, T3 is administered concurrently with an antidepressant from the inception of treatment and the outcome is examined after a treatment course of 4 wk or (preferably) more. Depending on the frequency of clinical evaluations in the initial stages, enhancement studies may also yield information on acceleration.

The thyroid axis is a classic example of an endocrine feedback loop. Hypothalamic thyrotropin-releasing hormone (TRH) stimulates the production of pituitary thyrotropin (thyroid-stimulating hormone, TSH) which in turn stimulates the thyroid gland to synthesize and secrete thyroid hormone. Thyroid hormone feeds back negatively to inhibit production of TRH and TSH. TSH level is the most sensitive and specific marker of thyroid function. The active thyroid hormones are T4 and T3. T4 is secreted from the thyroid gland in at least 20-fold excess over T3 yet T4 is considered to be a precursor for the more potent T3. Approximately 30% of T4 is converted to T3 by the deiodinase enzymes. Eighty percent of the T3 production and concentration in the blood comes from extra-glandular conversion of T4 to T3 and the remainder from thyroid secretion. T3 has approximately three times the hormonal potency of T4 due to greater bioavailability in the plasma and 10–15 times greater affinity for thyroid receptors. Thyroid hormones influence the growth and maturation of tissues, cell respiration and energy expenditure and the turnover of all substrates. Thyroid hormones act by binding to nuclear thyroid hormone receptors α and β and influence genomic expression.

The rationale for using thyroid hormone to supplement antidepressants is based on evidence supporting a connection between thyroid function and depression that has long been recognized but is still not well defined. Some (e.g. Gulseren et al., 2006; Thomsen et al., 2005; Wekking et al., 2005) but not all studies (Engum et al., 2002) support a link between hypothyroidism and mood disturbances, quality of life and neurocognitive functioning. These symptoms may improve as a result of treatment or remain unchanged despite T4 replacement (Gulseren et al., 2006; Wekking et al., 2005). Although the majority are euthyroid, patients with MDD have a higher than anticipated rate of subclinical hypothyroidism with high TSH (Brouwer et al., 2005; Fraser et al., 2004; Joffe and Levitt, 1992). In addition, depressed patients may show reduced (blunted) TSH response to the TRH stimulation test (Kastin et al., 1972; Prange et al., 1972) or reduced levels of the thyroid transport protein, transthyretin, in cerebrospinal fluid (Hatterer et al., 1993; Sullivan et al., 1999, 2006). There is evidence that an autoimmune process may be occurring in some depressed patients even in the absence of overt thyroid dysfunction. Higher thyroid-binding inhibitory immunoglobulin (TBII) and higher thyroid microsomal antibody (TMA) levels were found in all depressive subtypes compared to controls (Fountoulakis et al., 2004). However, in large population-based studies no associations were found between thyroid autoimmunity and depression (Engum et al., 2005; Grabe et al., 2005). Finally, partial substitution of the daily T4 allowance with T3 in patients with primary hypothyroidism was associated with improved mood and cognitive performance (Buneviĉius et al., 1999); this finding has been replicated only partially (Appelhof et al., 2005) or not at all (Clyde et al., 2003; Siegmund et al., 2004). A recent meta-analysis concluded that monotherapy with T4 remains the standard treatment for primary hypothyroidism and did not recommend a combination of T3 and T4 (Grozinsky-Glasberg et al., 2006).

Some studies report an association of clinical and sublinical hypothyroidism with reduced response to antidepressant treatment (Berlin et al., 1999; Corruble et al., 2004) but this is not consistently supported (Iosifescu et al., 2001; Joffe, 1999). Several changes have been reported in the thyroid function of depressed patients treated with antidepressants but their direction is not consistent (Gambi et al., 2005; Rao et al., 1996; Sagud et al., 2002). An association with response to treatment has been reported (Gendall et al., 2003; Shelton et al., 1993). Others attributed the changes in thyroid function to recovery rather than a direct effect of antidepressants (Duval et al., 1996). Overall, efficacy of thyroid hormones in depression may not be necessarily related to an underlying abnormality of thyroid function in these patients.

T4 has been studied less extensively than T3 in the treatment of affective disorders. This may be due to the fact that T4 is converted to the more potent T3, which is the active hormone. Nevertheless, there is evidence for efficacy of T4, particularly in patients with bipolar disorder. The addition of supraphysiological doses of T4 to antidepressant-resistant unipolar and bipolar depression, improved the outcome in two open clinical studies (Bauer et al., 1998; Pfeiffer et al., 2004). Supraphysiological doses of T4 also improved bipolar disorder prophylaxis in a prospective open study (Bauer et al., 2002b). In a recent open study, moderate doses of T4 successfully augmented serotonergic antidepressants in female patients with refractory unipolar or bipolar depression (Lojko and Rybakowski, 2007). Several studies support specific efficacy of T4 in the treatment of rapid-cycling bipolar disorder, reducing both amplitude and frequency of manic and depressive phases (Afflelou et al., 1997; Bauer and Whybrow, 1990; Whybrow, 1994). Finally, in one study comparing T4 and T3 potentiation of tricyclic antidepressants (TCAs), significantly more patients responded to T3 compared to T4 (Joffe and Singer, 1990). In one study supraphysiological doses of T4 were associated with serious adverse effects (Pfeiffer et al., 2004).

There is support from meta-analyses for augmentation and acceleration by T3 of the therapeutic effect of TCAs (Altshuler et al., 2001; Aronson et al., 1996). Aronson et al. (1996) aggregated eight studies (four of which were randomized, double-blind) that encompassed a total of 292 patients with treatment-resistant depression, in which T3 was added to ongoing treatment with TCAs. The term ‘treatment resistance’ was relative in that the patients received only one initial therapy with a TCA, given for a varying duration of 10 d to 12 wk. The T3 augmentation group was twice as likely to respond compared to controls (relative likelihood of response 2.09, 95% CI 1.31–3.32, p=0.002), corresponding to a 23.2% absolute improvement in response. The quality of the studies was uneven, the number of participants was relatively small, and the findings were not statistically significant when the analysis was restricted to double-blind studies. Nonetheless, the overall pattern suggested that T3 is an effective method of augmenting the antidepressant effects of TCAs in patients with treatment-resistant depression. The meta-analysis of Altshuler et al. (2001) focused on acceleration by T3 of the antidepressant effects of TCAs. In five out of six randomized, double-blind, controlled trials (RCTs) the average effect of T3 was highly significant compared to placebo in accelerating clinical response when administered concurrently with TCAs to patients with major depression (p=0.002). The effect was stronger in women (p=0.04).

There are relatively few studies of T3 as an accelerating, enhancing or augmenting agent in patients with MDD treated with specific serotonin reuptake inhibitors (SSRIs). The purpose of this review is to evaluate the published data from the perspective of efficacy and adverse effects and to suggest directions for future research. Since SSRIs are widely used first-line agents in the treatment of MDD and considering the low rates of response and remission achieved with SSRI monotherapy (Trivedi et al., 2006), this is an important clinical issue.


We searched PubMed for papers published since 1965 in English-language journals. The following main search components were combined using ‘and’ or ‘or’: (1) The different common terminologies for the thyroid hormone under study: ‘triiodothyronine’, ‘T3’, and ‘liothyronine,’ were used. (2) To identify the different types of studies the terms used were ‘open study’, ‘randomized controlled trial or study’, ‘RCT’ and ‘placebo’. (3) To identify the various uses of the hormone the terms used were: ‘supplementation’, ‘augmentation’, ‘acceleration’, ‘enhancement’, ‘combination’ and ‘potentiation’. (3) To identify depression, ‘major depressive disorder’, ‘MDD’, ‘major depression’ and ‘unipolar depression’ were used. (4) To identify treatment resistance the terms ‘treatment resistant depression’, ‘treatment refractory depression’, ‘TRD’, ‘resistance’ and ‘treatment failure’ were used. (5) The antidepressant medication was searched as: ‘antidepressants’, ‘antidepressant medication’, ‘specific serotonin reuptake inhibitors’, ‘SSRI’ and the generic names of all the SSRIs. Relevance was determined by screening the titles and abstracts. Reference lists of relevant articles were screened for additional potential articles. ‘Related articles’ appearing on any search were similarly screened. We supplemented the search in the Cochrane Collaboration Library.

The identified studies were divided into two groups: (1) studies that randomized patients to addition of T3 or addition of placebo or another agent and compared therapeutic and adverse effects, and; (2) studies that did not include randomization in which the effect of T3 was assessed without comparison to the effect of placebo or another agent. A second division was made between: (a) studies focusing on enhancement, where T3 was administered concurrently with an antidepressant from the inception of treatment and the therapeutic effect was assessed, and; (b) studies focusing on augmentation, where T3 was administered after the failure of a standard antidepressant trial. No studies were excluded. We evaluated and compared the results of these studies in order to draw conclusions regarding the degree to which addition of T3 leads to increased efficacy of a concurrently administered SSRI, the nature and severity of adverse effects associated with the intervention and possible predictors of response. The number of studies identified was insufficient in number and too variable in design to allow a meta-analysis to be performed. In the studies reviewed here response was consistently defined as a reduction of ⩾50% in intent-to-treat scores on the primary rating scale for depression. The criterion for remission was variable and is noted in the text.


Five studies were identified that included a randomization component (Appelhof et al., 2004; Cooper-Kazaz et al., 2007; Joffe et al., 2006; Nierenberg et al., 2006; Posternak et al., 2007). These are summarized in Table 1. Three were enhancement studies in which T3 and a SSRI were administered concurrently from the inception of treatment (Appelhof et al., 2004; Cooper-Kazaz et al., 2007; Posternak et al., 2007) and two were augmentation studies in which T3 was added to the ongoing treatment of patients who had not responded (Joffe et al., 2006; Nierenberg et al., 2006). No open enhancement studies were identified and no specifically designed acceleration studies. Three open augmentation studies were identified (Abraham et al., 2006; Agid and Lerer, 2003; Iosifescu et al., 2005). These are summarized in Table 2.

View this table:
Table 1

View this table:
Table 2

Enhancement studies

Appelhof et al. (2004) performed an 8-wk, randomized, double-blind, placebo-controlled trial of T3 added to paroxetine from the inception of treatment in outpatients with non-psychotic MDD. There were three treatment arms: paroxetine (up to 30 mg/d)+25 µg/d T3 (n=30), paroxetine+50 µg/d T3 (n=30) and paroxetine+placebo (n=53). The mean age of the participants was 46.5±11 yr and 62% were female. The three groups were similar at baseline with respect to demographic characteristics and severity and chronicity of depression. Mean baseline scores on the 17-item Hamilton Depression Rating Scale (HAMD17) were ∼21 for all groups. Endocrine measures (TSH, T3 and free T4) were similar and all patients were euthyroid at baseline. Of 113 patients who were randomized, 106 were included in the final analysis. There was a significant but similar improvement in the primary and secondary outcome measures on all three treatments. Overall rate of response was 46% in all the groups; the remission rate (HAMD17 ⩽8) was 32% in both T3 groups and 36% in the placebo group. No acceleration effect was found and there was no differential effect among female patients. Nine patients did not complete the study; one patient sought other treatment and eight patients dropped out due to side-effects, six of them within the first 2 wk of treatment. Although attrition rates were similar among the groups, patients treated with T3 experienced significantly more adverse events especially in the 50-µg group (palpitations, sweating, nervousness and tremor) than patients treated with placebo.

The study of Posternak et al. (2007) was a 6-wk, randomized, double-blind, placebo-controlled trial that sought to bridge the ‘efficacy-effectiveness gap’ by conducting a controlled trial under naturalistic conditions. Patients were recruited through an outpatient clinic. There were no psychiatric restrictions to participation; patients with bipolar disorder, MDD with psychotic features and other psychiatric comorbidity were not excluded. There were no restrictions on antidepressant or ancillary treatments. Randomization to T3 or placebo was at enrolment and was blinded. Fifty-seven patients were randomized and 50 were included in the final data analysis. The two treatment groups were antidepressant+25 µg/d T3 (n=23) and antidepressant+placebo (n=27). The mean age of the T3 group was 40±9.4 yr and of the placebo group 36±11.4 yr; 57% and 70% were female in the two groups respectively. Thirty percent of the T3 group and 26% of the placebo group had no prior adequate antidepressant trials. The average current episode duration was 51±80 months in the T3 group and 27±37 months in the placebo group. Sixty-five percent and 67% of the patients in the T3 and placebo groups respectively had an antidepressant newly initiated upon entering the study; 26% and 19% had their antidepressant switched due to non-response and 9% and 15% had an antidepressant switch due to relapse. Thus, the definition of this trial as an enhancement study is somewhat equivocal. Mean baseline scores on the Montgomery–Asberg Depression Rating Scale (MADRS) were 29.3±7.6 (T3) and 30.3±8.7 (placebo). Endocrine measures were similar for the two groups. Fifty-two percent of patients received a SSRI and the remaining 48% received other antidepressants (Table 1). Sixty-four percent of patients received one or more ancillary medication and 50% at least one session of psychotherapy during the study. The placebo group received significantly more psychotherapy sessions than the T3 group but there were no other differences in the other treatments received. The rate of completion was high in both groups (T3, 83%; placebo, 85%). Acceleration of response was found in the first 3 wk in the T3 group when continuous outcomes were analysed; the difference was significant at week 2. The dichotomous analysis found no acceleration. On the self-rated instrument a similar pattern was found but it did not achieve significance. With both scales a slight increase in depression severity was found at week 3. Response and remission (MADRS ⩽10) rates were numerically greater in the T3 group but the differences were not statistically significant. Hyperthyroidism-like side-effects were numerically greater in the placebo group but the only statistically significant difference was for nervousness.

Cooper-Kazaz et al. (2007) performed an 8-wk, randomized, double-blind, placebo-controlled trial of T3 administered concurrently with sertraline to outpatients with non-psychotic, non-refractory MDD. The treatment arms were sertraline (50 mg/d for the first week, 100 mg/d thereafter)+T3 (20–25 µg/d for the first week, 40–50 µg/d thereafter) (n=53) and sertraline+placebo (n=50). The two groups were similar at baseline with regard to demographic and clinical characteristics and thyroid measures. The mean ages were 45.15±58.00 yr and 40.98±55.77 yr in the T3 and placebo groups respectively and the percent of women was 55% and 58%. Only 8% of the patients had received an adequate antidepressant trial during the current episode (Stage I Resistant Depression; Thase and Rush, 1997); 11.6% were suffering from chronic depression. Baseline HAMD21 was 20.69±4.9 in the T3 group and 21.96±5.27 in the placebo group. Of the 124 randomized patients, 103 were included in the final analysis. The T3 group showed significant advantage over the placebo group in response (T3, 69.8%; placebo, 50.0%; p=0.02) and remission (HAMD21 ⩽6) rates (T3, 58.5%; placebo, 38.0%; p=0.02) as defined by HAMD21. No acceleration effect was found and there were no gender differences. Baseline T3 values were lower in patients treated with T3 who remitted compared to non-remitters (p<0.002). Among T3-treated patients, remission was associated with a significant decrease in serum TSH values (p<0.05). T3 supplementation was not associated with more frequent adverse effects.

Augmentation studies

The first open augmentation study of SSRIs by T3 was by Agid and Lerer (2003). They performed an algorithm-based study that included a SSRI treatment phase and the addition of T3 to non-responders. Ninety outpatients suffering from non-psychotic MDD commenced 4 wk of treatment with 20 mg/d of a SSRI [fluoxetine (n=81) or paroxetine (n=9)]. Seventy-four completed the 4 wk and of them 48.9% (n=44) were responders [Clinical Global Inventory B (CGI-B) score of at least 2 (much improved)]. The non-responders (n=30) had their SSRI dose increased to 40 mg/d for a further 2 wk. This step converted only 16.6% (n=5) to response. The non-responders to 6 wk of SSRI (n=25) were then administered 25 µg/d T3 for 2 wk while continuing their SSRI at the same dose. Four patients, who did not respond after 1 wk of 25 µg/d, had their T3 dose raised to 50 µg/d for one additional week of treatment. Concomitant treatment with benzodiazepines was allowed. T3 augmentation resulted in 40% (n=10) of the patients achieving response after 2 wk. According to the HAMD21 response criterion (HAMD21 reduction>50%) there were two additional responders. Five out of 15 non-responders to 25 µg/d T3, received 50 µg/d T3 for an additional (third) week but did not respond. Of the four whose T3 dose was raised to 50 µg/d after 1 wk, only one responded. Of special note is that all responders to T3 were female, i.e. 62.5% (10/16) of the women responded but none of the men (n=9). Although TSH values were within normal range, responders to T3 had significantly higher baseline TSH levels than non-responders. Responders were also characterized by more schooling, fewer hospitalizations and less family psychiatric history. The HAMD21 items associated with response to T3 were: agitation, somatic anxiety, genital symptoms, and diurnal variation. Side-effects were not mentioned.

Iosifescu et al. (2005) performed a 4-wk, open T3 augmentation study of patients suffering from non-psychotic MDD, resistant to an adequate trial of SSRI treatment (⩾8 wk with ⩾4 wk at a stable and sufficient dose). Patients had no medical contraindication to treatment with T3 and were euthyroid. All subjects received 50 µg/d T3 for 4 wk while continuing the pre-enrolment SSRI doses. After 4 wk of T3 augmentation, they continued in open follow-up for another 8 wk during which antidepressant changes were allowed. Twenty patients with a mean age of 44.3±10.3 yr were enrolled (11 female, 9 male). No patient had abnormal thyroid function. The length of the current episode was 35.1±46.9 months. T3 augmentation was associated with a statistically significant reduction in mean HAMD17 from a baseline score of 20.5±3.6 to a final score of 14.0±7.1 with a mean 32.1±32.9% improvement. In total, 35% (n=7) achieved response and 30% (n=6) achieved remission (final HAMD17 ⩽7). No differences in baseline thyroid function or baseline depression severity were found between responders and non-responders. Five subjects with atypical depression showed greater clinical improvement with higher response [100% for atypical (5/5) vs. 13.3% (2/15) for non-atypical] and remission rates [80% for atypical (4/5) vs. 13.3% (2/15) for non-atypical]. Contrary to atypical depression, eight patients with melancholic features experienced greater severity of their depression compared to non-melancholic, by the end of the study. No differences were found with respect to gender. After the 8 wk continuation phase all responders sustained their response and no non-responders became responders. Adverse events of fatigue and diaphoresis were reported by two subjects (10%) and tremor, dry mouth, headaches, muscle aches and vivid dreams by one subject (5%). Only one patient discontinued due to side-effects (not specified).

Abraham et al. (2006) performed a 3-wk minimum, open, T3 augmentation study of patients suffering from non-psychotic MDD who had not responded to ⩾6 wk of SSRI treatment. All patients were euthyroid, had no indication of an autoimmune thyroid disease and had a normal TRH stimulation test prior to inclusion. All patients continued taking the same SSRI at the same dose and received T3 (up to 50 µg/d) for an additional 3–4 wk. Twelve patients, eight female with a mean age of 45.1±18 yr and four male with a mean age of 52.3±16 yr were enrolled in the study. The patients were all treatment resistant; four in stage II, seven in stage III and one in stage IV (Thase and Rush, 1997). Two patients received olanzapine prior to and during the study. Women received a mean T3 dose of 40.6µg/d for 3.8 wk and men 43.8 µg/d for 3.5 wk. T3 augmentation was associated with a statistically significant reduction in HAMD17 from a mean baseline score of 27±4 to a final score of 17±8. Five patients (42%) achieved response and three of them achieved remission (final HAMD17 ⩽7). There were no differences between responders and non-responders in baseline HAMD17, number of previous antidepressant trials, gender or thyroid function. One patient withdrew during the first week of treatment due to increased anxiety and agitation. T3 was well tolerated by the other patients.

Nierenberg et al. (2006) performed a randomized, single-blind study that compared augmentation of ongoing antidepressant treatment by lithium (Li) and T3 in level 3 of the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study. Patients who entered the study had failed to remit on (or been intolerant of) treatment with citalopram in level 1 and in level 2 switched to bupropion, sertraline or venlafaxine or augmentation with buproprion or buspirone. Of the 4041 participants who enrolled in STAR*D, 142 entered the level 3 augmentation study. They were randomly assigned to Li (n=69) increased to 900 mg/d over a 2-wk period or T3 (n=73) increased to 50 µg/d over 1 wk. Augmentation was added to the ongoing antidepressant treatment (citalopram, bupropion-SR, sertraline or venlafaxine-XR) for up to 14 wk. The two previous augmenting options (bupropion-SR or buspirone) were discontinued at the initial visit. Concomitant treatment with anxiolytics (except alprazolam) or hypnotics was allowed. Mean ages were 43.2±11.8 yr and 40.6±12.2 yr in the T3 and Li groups respectively and the percent of women was 56.2% and 60.9%. A total of 26.8% of the T3 group and 23.9% of the Li group were suffering from chronic depression. Mean baseline score on HAMD17 was 17.2±6.2 for T3 and 19.0±6.6 for Li. The mean duration of augmentation was 9.9±5.0 wk for T3 and 9.3±5.4 for Li. Mean daily doses at exit were 859.8±373.1 mg/d for Li and 45.2±11.4 µg/d for T3. Median Li blood level was 0.6 mequiv/l. Remission (HAMD17 ⩽7) rates were 24.7% for T3 and 15.9% for Li; the difference was not statistically significant. On the self-reported secondary outcome measure (Quick Inventory of Depressive Symptoms – Self-Report; QIDS-SR16) there was no difference in response or remission rates. Mean times to response and remission were 5.7±5.1 wk and 7.4±4.4 wk in the Li group and 6.0±5.1 wk and 6.6±4.8 wk in the T3 group. Twenty-two percent of the Li remitters and 28% of T3 remitters did not reach remission until after 14 wk of treatment. Li was associated with more frequent side-effects and more side-effect-related exits from treatment (Li, 23.2%; T3, 9.6%, p=0.028).

Joffe et al. (2006) performed a 2-wk, randomized, double-blind, placebo-controlled, augmentation study of patients suffering from non-psychotic MDD who had not responded to a single, adequate trial of antidepressant treatment lasting at least 5 wk. Thirty-six patients were randomly assigned to one of four augmentation possibilities: 37.5 µg/d T3 (n=10), 600 mg/d Li (n=9), a combination of 37.5 µg/d T3 and 600 mg/d Li (n=9) or placebo (n=8). Li dosage was increased to 900 mg/d if after 1 wk of treatment the Li blood level was <0.6 mequiv/l. In total, 77.8% of patients were treated with a SSRI and the remainder with other antidepressants (Table 1). The antidepressant was maintained at the same dosage for the duration of the trial. All patients had normal thyroid function. Thirty-six patients were randomized: all completed the trial and all were included in the data analysis. Thirty of the patients were female and gender distribution among the groups was equal. The mean ages were 42.2±6.8 yr (T3), 38.3±6.8 yr (Li), 37.0±6.7 yr (T3+Li) and 38.8±7.4 yr (placebo). No data are available regarding specific features of the current depressive episode. The outcome measure was change in the HAMD17 score. The results showed a significant main effect of time, indicating an improvement in HAMD17 scores for all groups over the 2 wk. However, there was no significant treatment effect. Side-effects were not mentioned.


There are three pivotal issues to be addressed in evaluating these published studies of supplementation by T3 of SSRI treatment in MDD. The first is whether the intervention is efficacious. The second is how tolerable the intervention is. The third is whether any clinical insights can be gained as to which patients are more likely to respond, which doses should be administered and how long the treatment should be continued.


No studies were found that were specifically designed to examine acceleration of the antidepressant effect of SSRIs by T3. However, acceleration can be inferred from enhancement studies. Of the three enhancement studies reviewed, that of Posternak et al. (2007) offered weak support for acceleration by T3 of the antidepressant effect of SSRIs. Acceleration was not observed by Appelhof et al. (2004) or Cooper-Kazaz et al. (2007).

Three enhancement studies were reviewed, including two randomized. double-blind, placebo-controlled, efficacy studies (Appelhof et al., 2004; Cooper-Kazaz et al., 2007) and a single randomized, double-blind, placebo-controlled effectiveness study (Posternak et al., 2007). The two efficacy studies reached opposite conclusions – the results of Appelhof et al. (2004) did not support a role for T3 addition to SSRI whereas Cooper-Kazaz et al. (2007) showed significant enhancement of the effects of sertraline in terms of increased response and remission rates. The study of Posternak et al. (2007) found a non-significant trend for greater effectiveness of T3 than placebo. It is interesting to note that the results for the placebo arms of all three studies were comparable, response/remission rates being 46%/36% for Appelhof et al. (2004), 50%/38% for Cooper-Kazaz et al. (2007) and 52%/37% for Posternak et al. (2007). This could suggest that enhanced response and remission in the T3 arm is unique to the addition of T3, reflecting specific conditions or patient characteristics needed for T3 to have an effect. The Appelhof et al. (2004) and Cooper-Kazaz et al. (2007) studies were similar in inclusion/exclusion criteria, MDD recurrence rate, size of sample, depressive episode severity, length of study and T3 treatment. They differed in the SSRI administered (paroxetine vs. sertraline), antidepressant dose (30 mg/d paroxetine vs. 100 mg/d sertraline), chronicity of the depression (44% vs. 11.6%) and degree of resistance to antidepressant treatment (38%⩾Stage I vs. 8%). It is possible that a better response was observed in the Cooper-Kazaz et al. (2007) study because the patients were potentially more responsive to treatment and received a relatively lower antidepressant dose thus allowing more scope for a therapeutic effect of T3 to be demonstrated. It is more difficult to compare the results of the two efficacy studies with those of Posternak et al. (2007). Only 82% of the patients in the Posternak et al. (2007) study had non-psychotic, MDD, degree of chronicity is not mentioned, the length of the study was 6 wk and only 52% of the patients were treated with SSRIs. Further studies are needed to establish whether T3 is effective in enhancing the antidepressant effects of SSRIs.

Five augmentation studies were reviewed. The three small open studies (11, 19 and 25 patients) that examined the effect of T3 augmentation of SSRIs after non-response to 6–14 wk of treatment yielded encouraging results. Response/remission rates were 40%/36% (Agid and Lerer, 2003), 35%/30% (Iosifescu et al., 2005), 42%/25% (Abraham et al., 2006). T3 dose was up to 50 µg/d and the duration of augmentation treatment was 3–4 wk except in the Agid and Lerer (2003) study where it was 2 wk with non-responders going on to a third week. The two randomized augmentation studies are very different in several respects and difficult to compare. The STAR*D study (Nierenberg et al., 2006) was single-blind, lasted up to 14 wk and compared T3 and Li augmentation in 142 patients who failed to remit on (or were intolerant of) at least two antidepressants or one antidepressant plus an augmentation other than T3 or Li. In only 57.7% of patients was a SSRI augmented. Remission rates were 15.9% for Li and 24.7% for T3 but the difference was not statistically significant. In contrast, the Joffe et al. (2006) study was double-blind but lasted only 2 wk and involved only 36 patients randomized to four treatment groups rendering it clearly underpowered. A SSRI was augmented in 77.8% of the cases. Categorical response and remission rates were not given. There was no difference in effect of T3, Li, Li+T3 or placebo augmentation on HDRS17 scores. Overall, the data supporting efficacy of T3 augmentation of SSRIs in non-responsive patients are suggestive with remission rates of 25–36% but far from definitive. Further well-controlled trials are needed to establish whether T3 is an effective augmenting agent in patients treated with SSRIs who have not responded.

Adverse effects

Although T3 has been used for many years as a supplement to antidepressant medication and assessment of thyroid function is standard when treating patients suffering from depression, clinicians are still hesitant to use this agent. Their hesitation is at least partly due to reluctance to interfere with thyroid function in euthyroid individuals and concern regarding adverse effects. The potential adverse effects of T3 are tachycardia, nervousness, insomnia, tremor, diarrhoea, headache, angina, weight loss, diaphoresis and in severe cases arrhythmias or cardiac decompensation.

In contrast to these concerns, the studies reviewed here indicate a paucity of adverse effects associated with use of T3 as a supplement to SSRIs and other second-generation antidepressant agents. It is interesting to note that the study that showed the most T3- associated adverse effects also showed no therapeutic advantage to the use of T3 (Appelhof et al., 2004). In the other two double-blind, enhancement studies T3 was not associated with more adverse effects than placebo (Cooper-Kazaz et al., 2007; Posternak et al., 2007). Given the putative noradrenergic reuptake effects of paroxetine as well as the enhancing effects of thyroid hormone on adrenergic signalling (Frazer, 2001; Owens et al., 1997), it is possible that the higher adverse side-effect burden observed by Appelhof et al. (2004) might be specific to the combination of T3 with paroxetine. In the two open augmentation studies where information on adverse effects is given (Abraham et al., 2006; Iosefecu et al., 2005) T3 was well tolerated. In the Abraham et al. (2006) study one subject dropped out due to anxiety and agitation. In the Iosifescu et al. (2005) study two subjects developed adverse effects; one discontinued due to them. The adverse effects were fatigue and diaphoresis in both patients and tremor, dry mouth, headache, muscle pain and vivid dreams in one of them. These adverse effects can be attributed both to T3 and sertraline. In the STAR*D study (Nierenberg et al., 2006) adverse effects of Li were more common and disabling then T3 adverse effects and lead to higher attrition rates. We conclude, on the basis of the available literature, that T3 at the doses used (up to 50 µg/d) is safe and well tolerated by most patients; potential adverse effects should be monitored but should not deter clinicians from its use.

Clinical and laboratory correlates of response

Several of the studies reviewed reported clinical or laboratory correlates of response but these were not consistent. Agid and Lerer (2003) found a relationship between response to T3 and gender, all the responders in their study being female. In the meta-analysis of Althsuler et al. (2001) of acceleration effects of T3 when administered concurrently with TCAs, a gender effect was also observed, acceleration being more marked in female patients. However, none of the other studies reviewed here reported gender-related effects. Agid and Lerer (2003) also observed an association between response to T3 and thyroid function, responders to T3 having higher pretreatment TSH levels than non-responders but within the normal range. Cooper-Kazaz et al. (2007) observed an association of response to T3 and pre-treatment free T3 levels, which were lower in responders than non-responders. Moreover, responders to T3 had a greater reduction in TSH levels over the treatment course than non-responders, indicating a relationship of response to T3 to thyroid gland function. No relationship of response to thyroid function was observed by the other studies reviewed here. However, the thyroid measures used were not consistent between studies and the issue awaits clarification. Other correlates of response observed in specific studies should be noted since these were not consistently examined or reported. Iosifescu et al. (2005) observed a particularly good response to T3 in patients with atypical depression and a particularly poor response in patients with melancholic depression. Agid and Lerer (2003) found that responders to T3 were characterized by significantly more years of schooling, fewer previous psychiatric hospitalizations and less family history of psychiatric disorder and that baseline agitation, somatic anxiety, genital symptoms and diurnal variation were more pronounced in responders to T3. Considerably more work is needed in order to clarify whether there are any reliable clinical or laboratory predictors of response to T3 supplementation.

Additional questions for further study

It is unclear from the studies that were reviewed whether higher T3 dosage (40–50 µg) is more effective than lower dosages. In the one study where a formal comparison was made (Appelhof et al., 2004) neither the 25-µg nor the 50-µg doses were effective in enhancing paroxetine effects. As regards length of treatment, it is interesting to note that the open augmentation studies were short (2–4 wk) and yielded relatively high response rates. On the other hand, in the positive enhancement study of Cooper-Kazaz et al. (2007) patients converted to remission throughout the full 8 wk of treatment. In the STAR*D study (Nierenberg et al., 2006) 28% of the patients who remitted on T3 did so only after 14 wk of treatment. Thus, the findings of the two controlled studies that yielded positive results for T3 indicate that the effect of the intervention is not rapid and that longer treatment periods are needed than suggested by the open augmentation studies. It is also not clear how long T3 should be continued in patients who respond to it and whether there are adverse consequences of longer term administration. Until this issue is addressed empirically, it would be prudent not to continue T3 as a maintenance therapy in patients who respond to it but to taper and stop the hormone within the phase of continuation treatment. A further question that needs to be addressed is the timing of T3 supplementation. When T3 is administered as an enhancer from the outset of treatment with an antidepressant, approximately 30% of patients who are likely to remit on the antidepressant alone will be receiving the hormone unnecessarily. On the other hand, augmentation after ⩾8 wk of treatment might be too long a delay. Additional questions for further study include the relative efficacy of T4 and T3 as augmentation strategies and the problem of unrecognized bipolar disorder as a possible source of differential response (Rihmer and Aksikal, 2006; Shi et al., 2004).


Current evidence is not conclusive with regard to the efficacy of T3 either as an enhancer of antidepressant treatment with SSRIs or as an augmenter in patients who have not responded, although there is support for both effects. Further controlled studies are needed before it can be definitively established whether T3 is an effective supplement to SSRIs in patients with MDD. It is important that future studies employ uniform designs in order to allow them to be compared. T3 appears to be safe and well tolerated and adverse effects need not be an impediment to clinical use provided that thyroid function is monitored. Clinical and thyroid function correlates of response have been identified and these require further investigation as possible predictors of response. The appropriate timing of the T3 supplementation needs to be explored and the appropriate dose and length of treatment established.


Supported in part by grants (to R.C-K.) from the Professor Milton Rosenbaum Fund for Research in the Psychiatric Sciences and the National Institute for Psychobiology in Israel.

Statement of Interest

Professor Bernard Lerer is Editor-in-Chief of the International Journal of Neuropsychopharmacology. He was not involved in the review of this paper, which was conducted by a Field Editor outside the Editorial Office.


View Abstract