Biological Psychiatry Laboratory, Department of Psychiatry, Hadassah Hebrew University Medical Center, Jerusalem, Israel

Transcription

1 International Journal of Neuropsychopharmacology (2003), 6, Copyright f 2003 CINP DOI : /S X Algorithm-based treatment of major depression in an outpatient clinic: clinical correlates of response to a specific serotonin reuptake inhibitor and to triiodothyronine augmentation ARTICLE Ofer Agid and Bernard Lerer Biological Psychiatry Laboratory, Department of Psychiatry, Hadassah Hebrew University Medical Center, Jerusalem, Israel Abstract Clinicians who treat major depression are faced with a bewildering choice of antidepressants. Given that all have a lag period before they are effective, choice of the right antidepressant can potentially minimize the duration of symptoms, reduce cost and enhance compliance. Unfortunately, there is very little evidence upon which to base such a decision. We developed a treatment algorithm for non-psychotic, unipolar major depression, and applied it in our clinic. It includes progression from a specific serotonin reuptake inhibitor (SSRI), usually fluoxetine, from a dose of 20 mg to a dose of 40 mg in patients who do not respond. Non-responders to this dose receive augmentation with triiodothyronine (T3, mg). All interventions are for fixed time periods and guided by overall clinical improvement as defined by the Clinical Global Inventory. Ninety patients commenced open-label treatment with 20 mg SSRI (fluoxetine, n=81; paroxetine, n=9). Seventy-four patients completed 4 wk of treatment and 44 (48.9%) were responders (intent to treat analysis). Raising the SSRI dose to 40 mg for a further 2 wk was effective in only 5 patients (16.6%). Non-responders to SSRI were significantly more depressed at baseline as reflected by their rating scales scores. Addition of T3 was effective in 10 out of 16 women (62.5%) while none of the 9 males responded. Although values were within the normal range, patients who responded to T3 had higher serum thyroid-stimulating hormone levels than those who did not. Our experience with algorithmbased treatment of unipolar, non-psychotic major depression in outpatients suggests that more than 40% of patients will not respond to initial treatment with an SSRI even when the dose is increased to 40 mg/d; that severity of depression may be an important predictor of response and that T3 may be useful as an augmenter of response in SSRI non-responders but may be less effective in men than in women. The effect of T3 may be related to thyroid function even within the normal range. Received 4 August 2002; Reviewed 27 September 2002; Revised 27 October 2002; Accepted 30 October 2002 Key words: Algorithm, augmentation, fluoxetine, major depression, paroxetine, specific serotonin reuptake inhibitor, triiodothyronine. Introduction Algorithm-based treatment has become a useful tool for specialist and primary-care physicians facing an explosion of information and treatment possibilities and the rising costs of health care (Levichek et al., 2002; Pardasani et al., 2000; Teitel, 2000). The rationale for using an algorithm-based treatment approach for major depression lies in the large number of pharmacological treatments and treatment combinations for Address for correspondence : Dr O. Agid, CAMH, Clarke Division, 250 College St., PET Centre, Toronto, ON M5T 1R8, Canada. Tel. : (ext. 6412) Fax : oagid@camhpet.on.ca this common disabling illness. Treatment algorithms are based on evidence-based knowledge and expert consensus and offer a stepwise approach that simplifies the physician s treatment decisions. In recent years algorithms have been developed for the psychopharmacological treatment of unipolar depression (Trivedi et al., 1998), chronic depression (Trivedi and Kleiber, 2001), depressed primary-care patients (Mulsant et al., 2001), major depression in patients with advanced cancer (Nakano et al., 1999) and melancholia (Parker et al., 1995). Typically, an algorithm encompasses a decision tree with progression through the treatment stages according to the response of the patient.

2 42 O. Agid and B. Lerer More than 30 different antidepressants from six different psychopharmacological families are currently marketed. Except for some studies that suggest greater efficacy of tricyclic antidepressants compared to specific serotonin reuptake inhibitors (SSRIs) in severe or hospitalized depressives (Anderson, 1998; Danish University Antidepressant Group, 1986; Roose et al., 1994), there is little evidence indicating that one antidepressant drug or one psychopharmacological class is more effective than the others. There are a very large number of studies dealing with the broad subject of predictors of response to antidepressant drugs. These have not always differentiated between different types of depression (unipolar, bipolar, psychotic), examined a specific antidepressant agent, used rating scales to assess response or monitored the blood levels of the drug (for reviews see Bielski and Friedel, 1976; Goodwin, 1993; Joyce and Paykel, 1989; Schatzberg, 1998). Putative predictors include demographic and clinical characteristics, personality features, biological markers such as urinary and plasma monoamine metabolite levels, neuroendocrine markers such as the dexamethasone suppression test, sub-cortical hyperintensities seen on brain imaging and psychophysiological features such as shortened REM sleep latency (e.g. Dratcu and Calil, 1989; Ehlers et al., 1996; Schatzberg, 1998; Sotsky et al., 1991). In this paper we report data from the first two stages of an algorithm for the treatment of unipolar, non-psychotic major depression applied in our outpatient clinic an SSRI (usually fluoxetine) at standard and increased dosage and augmentation with triidothyronine (T3) and consider clinical predictors of response to these stages. Addition of T3 to ongoing tricyclic antidepressant treatment is considered to be an effective augmentation strategy in patients with refractory depression (Joffe, 1998). In a meta-analysis, Aronson et al. (1996) aggregated 8 studies, 4 of which were double blind, placebo controlled, with a total of 292 patients. Patients treated with T3 augmentation were twice as likely to respond as controls, a 23.2% absolute improvement in response rates. However, the effect was not significant when the analysis was limited to the 4 double-blind studies. In a subsequent meta-analysis, Altshuler et al. (2001) examined doubleblind, placebo-controlled studies assessing concomitant administration of T3 and an antidepressant to accelerate clinical response in patients with major depression. Five of the six studies that were identified found T3 to be significantly more effective than placebo in accelerating clinical response to tricyclic antidepressants and the meta-analysis revealed a significant overall effect. In addition, there are case reports of depressed patients who responded to T3 augmentation after failing to respond to an SSRI (Gupta et al., 1991; Joffe, 1992). Methods Patients All the patients were regular referrals to the Outpatient Clinic of the Department of Psychiatry at the Hadassah Hebrew University Medical Center, from October 1998 to May The treatment algorithm described below was applied in the clinic during this period by O.A., to whom patients with definite or suspected major depression were referred. Patients were treated in the context of the algorithm if the following conditions were fulfilled: (1) diagnosis of major depression: single episode or recurrent, according to DSM-IV criteria, with no history of manic or hypomanic periods during the current episode or in the past; (2) had not received antidepressant or antipsychotic medication during the current episode (benzodiazepines for sedation or sleep were not a contra-indication); (3) administration of a specific serotonin inhibitor as the first step in pharmacological treatment indicated according to clinical judgement (the same condition was applied to each further step in the algorithm); (4) no history or clinical evidence of thyroid dysfunction and, for patients receiving T3, serum thyroid-stimulating hormone (TSH) levels within the normal range; (5) no contra-indication to the drug or drug combination prescribed. It was explained to the patient that his/her treatment would be according to a treatment algorithm that was developed according to the consensus clinical judgement of the doctors working in the clinic and would be flexibly applied in accordance with the patient s clinical condition and preferences. Patients were not required to sign consent forms for the treatment itself but written informed consent was required for other research projects (e.g. single photon emission tomography scans) in which a number of the patients were asked to participate, as approved by the Helsinki Committee (Internal Review Board) of the Hadassah University Hospital. Diagnostic and rating instruments All patients underwent a clinical interview using the Structured Clinical Interview for Axis I DSM-IV Disorders (SCID) Hebrew version (Shalev et al., 1994). In addition, patients were administered the Hebrew University Depression Database Questionnaire (HUDD-Q) (Lerer et al., 1999) which covers personal

3 SSRI and T3 in major depression 43 background and demographic features; past and recent significant life events; psychiatric history including previous hospitalizations, treatment and suicide attempts; medical history and family history of mental illness in first-degree relatives. Prior to commencing treatment and weekly thereafter, patients were rated with the Clinical Global Inventory (CGI) (Guy, 1976) and the 21-item Hamilton Depression Scale (HAMD) (Hamilton, 1960) and were asked to rate their own condition by marking a 100 mm Visual Analogue Scale (VAS). Ratings were performed in person in about 80% of cases; if the patient was not seen in the clinic, he/she was contacted by telephone. The CGI and HAMD were completed and the patients were asked to mark the VAS at home and bring it to the next visit. Treatment algorithm The treatment algorithm was developed by consensus of clinicians who work in the department in the course of a series of meetings that were led by the authors. The consensus was based on the clinical experience and regular practice of the participants and took into account published guidelines for the treatment of major depression (American Psychiatric Association, 2000). The overriding principle of the algorithm is that it represents a structured application of clinical practice that would reasonably be applied in the absence of the imposed structure. For this reason, the algorithm was not regarded as a research protocol. The algorithm covers patients with unipolar and bipolar major depression and takes into account patients who are already in treatment when they are referred to the clinic as well as those who have not yet been treated during the current episode. Since this report covers only patients with unipolar major depression who had not received an antidepressant during the current episode, only the relevant steps of the algorithm are described here. The first step in the algorithm is treatment with a specific serotonin reuptake inhibitor (SSRI) at a dose of 20 mg/d. Although fluoxetine or paroxetine could be administered, fluoxetine was generally preferred for reasons of insurance coverage. Patients received 20 mg/d of the SSRI for up to 4 wk. If at this time, their condition was not significantly improved [CGI global improvement (CGI-B) sub-scale o3], the dose was increased to 40 mg/d for up to 2 wk (provided that there were no significant adverse effects). If, after receiving an SSRI for 6 wk (the last two at a 40 mg dose) the CGI-B rating was not >3, T3 was added at a dose of 25 mg/d. The T3 augmentation phase was scheduled to last a minimum of 2 wk. Patients, who did not manifest significant improvement (CGI-B >3) after this period, were then switched to a dual action drug (clomipramine or, if a tricyclic antidepressant was contra-indicated, venlafaxine). Since insufficient patients reached this stage of the algorithm in the period covered by this report, this phase is not covered here. Statistical analysis x 2 tests were used for categorical variables and Student s t test for continuous, normally distributed variables; p<0.05 (two tailed) was regarded as significant. All patients who attended at least one followup visit were included in the analysis (last observation carried forward LOCF). Values are given as mean standard deviation (S.D.). Results 20 mg SSRI During the period reported here, 90 patients (Table 1) commenced treatment with 20 mg SSRI (fluoxetine, n=81; paroxetine, n=9). Fifty-seven (63.3%) were female; their mean S.D. age was yr (range yr); 66 (73.3%) were married or living with a partner; 7 (7.7%) were recent immigrants, the rest were Israeli born or immigrated to Israel more than 5 yr previously; 15 (10.6%) had a primary-school education or less, 41 (45.6%) a secondary-school education and 34 (37.8%) at least 1 yr of higher education; 63 patients (70%) were employed or students, 23 (25.7%) were pensioners and 3 (3.3%) were unemployed. Of the 90 patients who commenced treatment with 20 mg SSRI, 74 completed 4 wk of treatment (fluoxetine, n=65; paroxetine, n=9); 16 were dropouts who returned for at least one visit after initiating treatment. Considering all 90 patients, with the ratings of those who dropped out after one visit carried forward, 44 (48.9%) were responders to SSRI (fluoxetine, n=40/ 81, 49.4%; paroxetine, n=5/9, 55.6%) according to a CGI-B rating of 2 (much improved) or 1 (very much improved) at week 4. Among the completers, the rate of response to 20 mg SSRI according to CGI-B was 44/74 (59.5%). Applying a response criterion of >50% reduction in HAMD score, the results were very similar (45/90, 50%; 41/74, 55.4%). The rate of remission (HAMD f7) among the 20 mg SSRI completers was 28/74 (37.8%). Among the 16 patients who did not complete the 20 mg SSRI phase, none fulfilled CGI-B or HAMD response criteria at their one assessment on 20 mg SSRI.

4 44 O. Agid and B. Lerer Table 1. Patient numbers for each of the stages of the algorithm n Responders n (%) Non-responders n (%) 20 mg SSRI All patients (LOCF*) (48.9) 46 (51.1) Completers (59.5) 30 (40.5) 40 mg SSRI All patients/completers 30 5 (16.7) 25 (83.3) T3 All patients/completers (40) 15 (60) Response was defined according to CGI-B as indicated in the text. * LOCF, last observation carried forward. Table 2. Background and clinical characteristics of responders and non-responders to 20 mg SSRI for 4 wk Responders (n=44) Mean (S.D.) or n (%) Non-responders (n=30) Mean (S.D.) or n (%) There was no significant difference in age, sex or marital status between responders or non-responders to 20 mg SSRI who completed 4 wk of treatment but responders had slightly more years of schooling (p=0.04, Table 2), had experienced an adverse life event more frequently during the 6 months preceding the onset of their depression (p=0.01) and had been hospitalized for depression less frequently (p=0.04). Non-responders were significantly more depressed at baseline as reflected by their HAMD total (p= ), CGI-A (p=0.0002) and VAS (p=0.0002) scores. The HAMD items significantly associated with response were late insomnia (responders, ; Significance (x 2 test a or Student s t test b ) Age (yr) (15.1) 51.8 (17.8) ns Sex (male) 16 (36.6) 11 (36.7) ns Marital status (married) 32 (72.7) 21 (70.0) ns Years of schooling p=0.03 b Early parental loss 12 (31.6) 10 (37.0) ns Adverse life event 21 (47.7) 6 (20.0) p=0.01 a Previously hospitalized 6 (13.6) 10 (33.3) p=0.04 a First depressive episode 19 (43.2) 10 (33.3) ns Age of onset (yr) ns Suicide attempt 3 (6.8) 3 (10) ns Melancholic features 14 (31.8) 13 (43.3) ns Duration of episode (wk) ns Family history (affective) 11 (25.6) 13 (43.3) ns Family history (any) 14 (32.6) 15 (50.0) ns HAMD total p= b CGI-A p= b VAS p = b non-responders, ; p=0.008), psychic anxiety (responders, ; non-responders, ; p=0.0007), somatic symptoms and appetite (responders, ; non-responders, ; p=0.04) and diurnal variation (responders, ; non-responders, ; p=0.04). 40 mg SSRI Thirty patients, who completed 4 wk of treatment with 20 mg SSRI but did not respond, went on to receive 40 mg SSRI (fluoxetine, n=26; paroxetine, n=4). All completed at least 2 wk at this dose. Only 5 patients

5 SSRI and T3 in major depression 45 Table 3. Background and clinical characteristics of responders and non-responders to 40 mg SSRI mg T3 for 2 wk Responders (n=10) Mean (S.D.) or n (%) Non-responders (n=15) Mean (S.D.) or n (%) Significance (x 2 test a or Student s t test b ) Sex Female 10 (62.5) 6 (37.5) p=0.002 a Male 0 (0) 9 (100) Marital status (married) 7 (70) 10 (66.7) ns Years of schooling p=0.03 b Early parental loss 4 (40) 6 (46.2) ns Recent adverse life event 1 (10) 3 (20) ns Previously hospitalized 1 (10) 9 (60) p=0.01 a First depressive episode 4 (40) 3 (20) ns Age of onset (yr) ns Suicide attempt 1 (10) 2 (13.3) ns Melancholic features 4 (40) 9 (60) ns Duration of episode (wk) ns Family history (affective) 3 (30) 9 (60) ns Family history (any) 3 (30) 11 (73.3) p=0.03 a HAMD total ns CGI-A ns VAS ns TSH (mu/ml)* 1.88 (0.37) 1.46 (0.36) p=0.01 b * Responders, n=9; non-responders, n=14; normal range, mu/ml. (16.6%) responded to the higher SSRI dose, according to CGI-B >3. 3 (10%) were remitters, with HAMD scores f7. There were 2 additional patients who had HAMD change scores of 50 and 53%, respectively. Since decisions within the algorithm were based on CGI-B and these 2 patients were not sufficiently improved on this criterion, they went on to the T3 stage. Non-responders to 40 mg SSRI had significantly higher HAMD total scores (responders, ; non-responders, ; p= ), CGI-A scores (responders, ; non-responders, ; p=0.003) and lower VAS scores (responders, ; non-responders, ; p=0.0006). The HAMD items significantly associated with response were psychic anxiety (responders, ; nonresponders, ; p=0.03), appetite (responders, ; non-responders, ; p=0.002) and genital symptoms (responders, ; nonresponders, ; p=0.04). T3 All 25 patients who did not respond to 40 mg SSRI went on to receive augmentation with T3 (25 mg/d) for at least 2 wk. The T3 dose of 4 patients was raised to 50 mg/d after 1 wk. After 2 wk, 10 patients (40%) fulfilled the CGI-B response criterion (>3); 9 of them were remitted with a HAMD score of f7. According to the HAMD, there were 2 additional responders (50 and 53% HAMD change scores, respectively) who did not fulfil the CGI response criterion. It should be pointed out that of the 15 CGI-B non-responders to T3 after 2 wk, 5 received T3 for at least 1 wk more (i.e. 3 wk in all) but did not respond. Of the 4 patients whose T3 dose was raised to 50 mg/d after 1 wk, only one was a responder. Comparing responders and non-responders to T3, the most striking difference is in the sex of the patients. Ten of the 16 females responded to T3 (62.5%) while none of the 9 males responded (Table 3). Although values were all within the normal range ( mu/l), responders to T3 had significantly higher serum TSH levels than non-responders (p=0.01). This was also true when limiting the comparison to female patients (responders, n=9, mu/ml; nonresponders, n=6, mu/ml; p=0.008). The responders were also characterized by significantly more years of schooling, fewer previous hospitalizations and less family history of psychiatric disorder. The HAMD items significantly associated with

6 46 O. Agid and B. Lerer response to T3 were agitation (responders, ; non-responders, ; p=0.0005), somatic anxiety (responders, ; non-responders, ; p=0.03), genital symptoms (responders, ; non-responders, ; p=0.03) and diurnal variation (responders, ; nonresponders, ; p=0.001). Among the 5 patients whose T3 dose was raised to 50 mg/d after 1 wk, only one was a responder. Discussion Applying an algorithm-based protocol to the treatment of outpatients with unipolar, non-psychotic major depression who had not received an antidepressant during their current episode, we found that 48.9% were responders after 4 wk of treatment with an SSRI. According to the criterion we applied this meant that they were at least much improved on assessment with the CGI. Based on a >50% reduction in HAMD scores from baseline the response rate was 50%. These results were based on a conservative intent to treat analysis that included patients who did not receive the full duration of treatment. Considering only those patients who actually completed 4 wk of treatment, the proportion of responders according to CGI-B was 59.5%. Continuation of treatment for an additional 2 wk at double the initial dose was effective in 16.6% of patients who were non-responders at 4 wk giving an overall response rate of 54.4% after 6 wk of SSRI treatment. Focusing on fluoxetine, which was the primary drug administered to our patients, and considering a response criterion of 50% improvement from baseline, there is considerable variability in the literature regarding response to treatment. Findings range from 67% (Fava et al., 1995) through 55.5% (Nierenberg et al., 2000) to 51% (Flament et al., 1999). Overall, our findings support the widely accepted view that antidepressant treatment, while effective compared to placebo in controlled trials, is of therapeutic benefit to 60% or fewer patients in the regular clinical context (Sackeim, 2001). While algorithm-based treatment is a useful strategy in clinical practice, limitations of the approach in the research context should be borne in mind. A particularly relevant consideration relates to the later stages of the algorithm when effects observed may be related to the natural course of the illness. Furthermore, as the algorithm progresses, patients who do not respond, represent an increasingly selected population. Consequently, results with a particular intervention may not be representative of what might be observed in a situation where the particular treatment is used in a first-line context. On the other hand, the results of algorithm-based treatment more closely approximate actual clinical practice than do clinical trials. In view of the relatively low response rate that we and others have observed, variables that might to be used to predict response to SSRI treatment in the clinical context are of great potential importance. In the patient group that we studied, severity of depression was associated with poorer response to both the 20 and 40 mg SSRI doses as reflected by the observer and self-rated scales that were administered. Lesser baseline severity of depression has been associated with better response to fluoxetine in other studies (Klieser et al., 1995; Tedlow et al., 1998). Fewer hospitalizations (due to depression), which we found to be a positive predictor, could be connected to the positive predicting factor of lesser severity of depression. Similarly, the greater frequency of adverse life events in the SSRI responders could suggest a more situational than endogenous type of depressive background. Responders to SSRI in our sample had less anxiety at baseline; in other studies the impact of baseline anxiety levels on response has been variable (Burns et al., 1995; Klieser et al., 1995; Spillmann et al., 1997; Tedlow et al., 1998). Other variables reported to be associated with response include older age, greater weight and level of education (Berlin and Lavergne, 1997), early response to treatment (Koran et al., 1995) and baseline weight loss (Burns et al., 1995). Reduced insight and a tendency to blame others significantly predicted non-response (Spillmann et al., 1997). Recently, attention has been focused on the potential role of genetic variation in the serotonin transporter protein as a possible predictor of response to SSRIs. The data supporting this pharmacogenetic association and the potential role of other genes have recently been extensively reviewed (Lerer and Macciardi, 2002). As noted, increasing the SSRI dose from 20 to 40 mg/d was not particularly effective in our sample, converting only 5 out of 30 non-responders at 4 wk to responders by 6 wk. Although conclusions from this stage are limited because of the small sample size, one may question the usefulness of dosage increase. Rickels et al. (1985) compared fluoxetine in two daily dosage regimens (20 and 40 mg) in major depression and found no significant differences in efficacy between the two groups. Benkert et al. (1997) studied depressed outpatients who received either 20 mg paroxetine or 100 mg maprotiline for 3 wk; those patients with inadequate treatment response were randomized to either continuation of the previous dosage or dose

7 SSRI and T3 in major depression 47 increase. No significant benefit of dose increase was found for either drug. One possible explanation for this observation could be the percentage of serotonin transporter (5-HTT) sites occupied. In a recent PET study (Meyer et al., 2001) it was shown that for patients treated with 20 mg/d of paroxetine the mean proportion of 5-HTT sites occupied was 83%, 5-HTT occupancy increased in a nonlinear relationship with serum levels of paroxetine such that a plateau of occupancy around 85% occurred for serum paroxetine levels greater than 28 mg/l. On the other hand, Fava et al. (1995) found that increasing the dose of fluoxetine to 40 mg/d was effective in the treatment of relapse among depressed patients who had initially responded to 20 mg/d fluoxetine. All 25 of our patients who did not respond to 40 mg SSRI, went on to receive augmentation with T3. The association of gender with response was striking; 62.5% of the female patients responded to T3 while none of the males responded. This finding is of considerable interest but should be regarded with caution because of the small number of subjects. In their metaanalysis, Altshuler et al. (2001) found that the effect of T3 to accelerate antidepressant response was significantly related to the percentage of women included in the study samples, as previously suggested by Prange (1996). Aronson et al. (1996) did not include this perspective in their meta-analysis of augmentation studies with T3. It is noteworthy that in our report of 4 patients with post-traumatic stress disorder who responded to augmentation of SSRI with T3, 3 were male (Agid et al., 2001). Also, Stern et al. (1991) found a significant effect of T3 was to enhance response to ECT in male patients. Altshuler et al. (2001) speculated that the effect of gender on acceleration of antidepressant response by T3 might be related to a greater prevalence of clinical or sub-clinical thyroid dysfunction in women. Patients with known thyroid dysfunction were excluded from our sample and serum TSH levels for all the patients administered T3 were within normal limits. Nevertheless, the patients who responded to T3 had significantly higher serum TSH levels than those who did not although it should be noted that the difference was small and within the range of possible assay variation. This intriguing observation suggests that the relationship between thyroid function and response to T3 in patients with major depression needs to be explored further. In keeping with this suggestion, a subtle relationship between thyroid function, even within the normal range, and response to antidepressant treatment was recently demonstrated by Berlin et al. (1999) in patients with major depression. The mechanism whereby T3 may potentiate or augment the therapeutic effect antidepressant drugs is not definitively known. Recently, attention has been focused on a possible enhancement of serotonergic neurotransmission (see Bauer et al., 2002; Newman et al., 2000 for reviews). In-vivo microdialysis studies by our group (Gur et al., 1999) have shown that chronic administration of T3 or T3+clomipramine to rats results in an increase in cortical 5-HT concentrations and a desensitization of autoinhibitory 5-HT 1A receptors in the raphe area. These receptors control 5-HT release by a negative feedback mechanism. It is noteworthy that animals which received both T3 and clomipramine injections had basal 5-HT levels in the frontal cortex which were significantly higher than those seen in animals given either T3 or clomipramine alone (Gur et al., 1999). An increase in the sensitivity of cortical 5-HT 2 receptors by T3 has also been reported (reviewed by Bauer et al., 2002). Overall, our experience with algorithm-based treatment of unipolar, non-psychotic major depression in outpatients suggest that (i) more than 40% of patients will not respond to initial treatment with an SSRI even when the dose is increased to 40 g/d; (ii) severity of depression may be an important predictor of response; (iii) T3 may be useful as an augmenter of response in SSRI non-responders, although it may be less effective in men than in women. The effect of T3 to augment SSRIs may be related to thyroid function even within the normal range. Acknowledgements The Biological Psychiatry Unit, Hadassah Hebrew University Medical Center, is supported by the Harry Stern Family Foundation. References Agid O, Shalev AY, Lerer B (2001). Triiodothyronine augmentation of selective serotonin reuptake inhibitors in posttraumatic stress disorder. Journal of Clinical Psychiatry 62, Altshuler LL, Bauer M, Frye MA, Gitlin MJ, Mintz J, Szuba MP, Leight KL, Whybrow PC (2001). Does thyroid supplementation accelerate tricyclic antidepressant response? A review and meta-analysis of the literature. American Journal of Psychiatry 158, American Psychiatric Association (2000). Practice guideline for the treatment of patients with major depressive disorder. American Journal of Psychiatry 157, Anderson IM (1998). SSRIS versus tricyclic antidepressants in depressed inpatients : a meta-analysis of efficacy and tolerability. Depression and Anxiety 7,

8 48 O. Agid and B. Lerer Aronson R, Offman HJ, Joffe RT, Naylor CD (1996). Triiodothyronine augmentation in the treatment of refractory depression. A meta-analysis. Archives of General Psychiatry 53, Bauer M, Heinz A, Whybrow PC (2002). Thyroid hormones, serotonin and mood: of synergy and significance in the adult brain. Molecular Psychiatry 7, Benkert O, Szegedi A, Wetzel H, Staab HJ, Meister W, Philipp M (1997). Dose escalation vs. continued doses of paroxetine and maprotiline: a prospective study in depressed out-patients with inadequate treatment response. Acta Psychiatrica Scandinavica 95, Berlin I, Lavergne F (1997). Early predictors of two months response with mianserin and SSRI and influence of definition of outcome on prediction. European Psychiatry 13, Berlin I, Payan C, Corruble E, Puech AJ (1999). Serum thyroid-stimulating-hormone concentration as an index of severity of major depression. International Journal of Neuropsychopharmacology 2, Bielski RJ, Friedel RO (1976). Prediction of tricyclic antidepressant response : a critical review. Archives of General Psychiatry 33, Burns RA, Lock T, Edwards DR, Katona CL, Harrison DA, Robertson MM, Nairac B, Abou-Saleh MT (1995). Predictors of response to amine-specific antidepressants. Journal of Affective Disorders 35, Danish University Antidepressant Group (1986). Citalopram: clinical effect profile in comparison with clomipramine. A controlled multicenter study. Psychopharmacology 90, Dratcu L, Calil HM (1989). The dexamethasone suppression test: its relationship to diagnoses, severity of depression and response to treatment. Progress in Neuropsychopharmacology and Biological Psychiatry 13, Ehlers CL, Havstad JW, Kupfer DJ (1996). Estimation of the time course of slow-wave sleep over the night in depressed patients: effects of clomipramine and clinical response. Biological Psychiatry 39, Fava M, Rappe SM, Pava JA, Nierenberg AA, Alpert JE, Rosenbaum JF (1995). Relapse in patients on long-term fluoxetine treatment: response to increased fluoxetine dose. Journal of Clinical Psychiatry 56, Flament MF, Lane RM, Zhu R, Ying Z (1999). Predictors of an acute antidepressant response to fluoxetine and sertraline. International Clinical Psychopharmacology 14, Goodwin FK (1993). Predictors of antidepressant response. Bulletin of the Menninger Clinic 57, Gupta S, Masand P, Tanquary JF (1991). Thyroid hormone supplementation of fluoxetine in the treatment of major depression. British Journal of Psychiatry 159, Gur E, Lerer B, Newman ME (1999). Chronic clomipramine and triiodothyronine increase 5-HT levels in rat frontal cortex in vivo: relationship to 5-HT autoreceptor activity. Journal of Pharmacology and Experimental Therapeutics 288, Guy W (1976). ECDEU Assessment Manual for Psychopharmacology. NIMH. Hamilton M (1960). A rating scale for depression. Journal of Neurology Neurosurgery and Psychiatry 23, Joffe RT (1992). Triiodothyronine potentiation of fluoxetine in depressed patients. Canadian Journal of Psychiatry 37, Joffe RT (1998). The use of thyroid supplements to augment antidepressant medication. Journal of Clinical Psychiatry 59, 26 29; discussion Joyce PR, Paykel ES (1989). Predictors of drug response in depression. Archives of General Psychiatry 46, Klieser E, Lehmann E, Heinrich K (1995). Fluoxetine in patients with major depressive disorder a responder analysis. Pharmacopsychiatry 28, Koran LM, Hamilton SH, Hertzman M, Meyers BS, Halaris AE, Tollefson GD, Downs JM, Folks DG, Jeste DV, Lazarus LW, Satlin A. (1995). Predicting response to fluoxetine in geriatric patients with major depression. Journal of Clinical Psychopharmacology 15, Lerer B, Macciardi M (2002). Pharmacogenetics of antidepressant and mood stabilising drugs: a review of candidate gene studies and future research directions. International Journal of Neuropsychopharmacology 5, Lerer B, Shapira B, Bloch M, Hanin B, Trudart T, Alexander JR, Popper M, Braun D, Segman RH, Ritsner M (1999). Possible precipitants of psychiatric hospitalization in patients with major depression: results from the Jerusalem Collaborative Depression Project. Depression and Anxiety 9, Levichek Z, Atanackovic G, Oepkes D, Maltepe C, Einarson A, Magee L, Koren G (2002). Nausea and vomiting of pregnancy. Evidence-based treatment algorithm. Canadian Family Physician 48, , 277. Meyer JH, Wilson AA, Ginovart N, Goulding V, Hussey D, Hood K, Houle S (2001). Occupancy of serotonin transporters by paroxetine and citalopram during treatment of depression: a [( 11 )C]DASB PET imaging study. American Journal of Psychiatry 158, Mulsant BH, Alexopoulos GS, Reynolds 3rd CF, Katz IR, Abrams R, Oslin D, Schulberg HC (2001). Pharmacological treatment of depression in older primary care patients: the PROSPECT algorithm. International Journal of Geriatric Psychiatry 16, Nakano T, Kugaya A, Akechi T, Mikami I, Okuyama T, Okamura H, Uchitomi Y (1999). Algorithm for the treatment of major depression in patients with advanced cancer. Psychiatry and Clinical Neurosciences 53 (Suppl.), S61 S65. Newman ME, Agid O, Gur, E, Lerer B (2000). Pharmacological mechanisms of T3 augmentation of antidepressant action. International Journal of Neuropsychopharmacology 3, Nierenberg AA, Farabaugh AH, Alpert JE, Gordon J, Worthington JJ, Rosenbaum JF, Fava M (2000). Timing of onset of antidepressant response with fluoxetine treatment. American Journal of Psychiatry 157,

9 SSRI and T3 in major depression 49 Pardasani AG, Feldman SR, Clark AR (2000). Treatment of psoriasis : an algorithm-based approach for primary care physicians. American Family Physician 61, , 736. Parker G, Hadzi-Pavlovic D, Hickie I, Brodaty H, Boyce P, Mitchell P, Wilhelm K (1995). Sub-typing depression. III. Development of a clinical algorithm for melancholia and comparison with other diagnostic measures. Psychological Medicine 25, Prange Jr. AJ (1996). Novel uses of thyroid hormones in patients with affective disorders. Thyroid 6, Rickels K, Smith WT, Glaudin V, Amsterdam JB, Weise C, Settle GP (1985). Comparison of two dosage regimens of fluoxetine in major depression. Journal of Clinical Psychiatry 46, Roose SP, Glassman AH, Attia E, Woodring S (1994). Comparative efficacy of selective serotonin reuptake inhibitors and tricyclics in the treatment of melancholia. American Journal of Psychiatry 151, Sackeim HA (2001). The definition and meaning of treatment-resistant depression. Journal of Clinical Psychiatry 62, Schatzberg AF (1998). Noradrenergic versus serotonergic antidepressants: predictors of treatment response. Journal of Clinical Psychiatry 59, Shalev A, Abramowitz M, Kaplan A (1994). Hebrew edition of the Structured clinical interview for Axis I DSM IV patient edition. Spitzer RL, Williams JBA, Gibbon M (Eds.) New York, NY: New York State Psychiatric Institute: Sotsky SM, Glass DR, Shea MT, Pilkonis PA, Collins JF, Elkin I, Watkins JT, Imber SD, Leber WR, Moyer J (1991). Patient predictors of response to psychotherapy and pharmacotherapy: findings in the NIMH Treatment of Depression Collaborative Research Program. American Journal of Psychiatry 148, Spillmann M, Borus JS, Davidson KG, Worthington 3rd JJ, Tedlow JR, Fava M (1997). Sociodemographic predictors of response to antidepressant treatment. International Journal of Psychiatry Medicine 27, Stern RA, Nevels CT, Shelhorse ME, Prohaska ML, Mason GA, Prange Jr. AJ (1991). Antidepressant and memory effects of combined thyroid hormone treatment and electroconvulsive therapy: preliminary findings. Biological Psychiatry 30, Tedlow J, Fava M, Uebelacker L, Nierenberg AA, Alpert JE, Rosenbaum J (1998). Outcome definitions and predictors in depression. Psychotherapy and Psychosomatics 67, Teitel JM (2000). Unexpected bleeding disorders: algorithm for approach to therapy. Clinical Laboratory Haematology 22 (Suppl. 1), 26 9; discussion, Trivedi MH, DeBattista C, Fawcett J, Nelson C, Osser DN, Stein D, Jobson K (1998). Developing treatment algorithms for unipolar depression in cyberspace : International Psychopharmacology Algorithm Project (IPAP). Psychopharmacology Bulletin 34, Trivedi MH, Kleiber BA (2001). Algorithm for the treatment of chronic depression. Journal of Clinical Psychiatry 62,