- Email: [email protected]
The early non-increase of serum BDNF predicts failure of antidepressant treatment in patients with major depression: A pilot study
Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
Contents lists available at ScienceDirect
Progress in Neuro-Psychopharmacology & Biological Psychiatry j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p n p
The early non-increase of serum BDNF predicts failure of antidepressant treatment in patients with major depression: A pilot study André Tadić a,⁎,1, Stefanie Wagner a,1, Konrad Friedrich Schlicht a, Dirk Peetz b, Liudmyla Borysenko a, Nadine Dreimüller a, Christoph Hiemke a, Klaus Lieb a a b
Department of Psychiatry and Psychotherapy, University Medical Centre Mainz, Germany Institute for Clinical Chemistry and Laboratory Medicine, University Medical Centre Mainz, Germany
a r t i c l e
i n f o
Article history: Received 28 June 2010 Received in revised form 9 August 2010 Accepted 15 August 2010 Available online 20 August 2010 Keywords: Biomarker Brain-derived neurotrophic factor Major depressive disorder Serum Treatment outcome
a b s t r a c t In the treatment of patients with major depressive disorder (MDD), early non-improvement of symptoms after initiation of antidepressant treatment is a highly sensitive and specific marker for final treatment failure. On the other hand, meta-analyses of clinical studies investigating serum BDNF (sBDNF) concentration before and after antidepressant treatment showed an increase of sBDNF during treatment, which was correlated with amelioration of depressive symptoms. No study has yet investigated the predictive value of early changes of sBDNF for final treatment outcome of the individual patient. The aim of this study was to investigate in patients with MDD, whether i) the non-increase of sBDNF in the early course of treatment is a specific and sensitive marker for final treatment failure, ii) whether the sensitivity and specificity of early non-improvement for treatment failure can be increased by combining it with the marker “early nonincrease of sBDNF”. For this purpose, we performed a pilot study with 41 inpatients with MDD according to DSM-IV, who were treated in a naturalistic setting. Depression severity and sBDNF were measured in weekly intervals from baseline to week six with the 21-item Hamilton Depression Rating Scale (HAMD-21) and ELISA, respectively. The individual markers sBDNF non-increase and HAMD-21 non-improvement from baseline to day 7 or 14 predicted later non-response and non-remission with moderate to high specificity. The combined marker sBDNF non-increase plus HAMD-21 non-improvement at day 14 increased the specificity for non-response and non-remission to 100%. Our data provide the first evidence that the absence of an early increase of sBDNF in conjunction with early non-improvement might be a highly specific peripheral marker predictive for treatment failure in patients with MDD. If replicated, this combined marker could be considered useful for prospective confirmatory trials in patients with MDD. © 2010 Elsevier Inc. All rights reserved.
1. Introduction In the treatment of Major Depressive Disorder (MDD) outcomes with currently available antidepressants are disappointing, because only 30–40% of “real-life” patients with MDD experience a remission Abbreviations: AD, antidepressant; BL, baseline; D7, day 7; D14, day 14; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders; ELISA, enzyme-linked immunosorbent assay; EP, endpoint; HAMD-21, 21-item version of the Hamilton Depression Rating Scale; ITT, intention-to-treat; MDD, major depressive disorder; NPV, negative predictive value; PASW, Predictive Analysis SoftWare; PPV, positive predictive value; sBDNF, serum brain-derived neurotrophic factor; SD, standard deviation. ⁎ Corresponding author. Department of Psychiatry and Psychotherapy, University Medical Centre of the Johannes-Gutenberg-University Mainz, Untere Zahlbacher Str. 8, Germany. Tel.: + 49 6131 17 3950; fax: + 49 6131 17 3459. E-mail addresses: [email protected] (A. Tadić), [email protected] (S. Wagner), [email protected] (K.F. Schlicht), [email protected] (D. Peetz), [email protected] (L. Borysenko), [email protected] (N. Dreimüller), [email protected] (C. Hiemke), [email protected] (K. Lieb). 1 These authors contributed equally to this manuscript. 0278-5846/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.pnpbp.2010.08.011
with the first selected antidepressant (Warden et al., 2007, and refs. inside). Their effect sizes are rather small than medium (Kirsch et al., 2008; Turner et al., 2008; Fournier et al., 2010) and more effective antidepressant substances cannot be expected in the near future. Therefore, it is indispensable to develop new strategies to increase remission rates and to shorten the time to remission in acutely depressed patients. One approach is the identification of individual patients who will not respond to a selected antidepressant treatment as early as possible and a consecutive treatment change in these patients. A substantial body of evidence from many retrospective studies incorporating in total more than 12,000 patients and virtually all groups of antidepressants (Henkel et al., 2009; Hennings et al., 2009; Katz et al., 2004; Nierenberg et al., 2000; Stassen et al., 2007; Szegedi et al., 2003, 2009; Tadić et al., 2010a) shows that non-improvement, oftenly defined as b20% reduction of depressive symptoms (assessed with a routine scale like the Hamilton Depression Rating Scale (HAMD)) after 14 days of treatment is a highly specific marker for final treatment failure with an unchanged treatment. E.g., the largest
416
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
meta-analysis in this field (Szegedi et al., 2009) showed that nonimprovement after two weeks of treatment is a highly specific predictor for final non-response (81–98% across the investigated studies) and non-remission (87–100%). Based on these data, the application of the marker early non-improvement in clinical routine might be able to reduce the treatment time with an insufficient drug to two weeks. Its potential to increase remission rates in the treatment of acutely depressed patients is currently investigated by a large randomized clinical trial (Tadić et al., 2010b). The above mentioned figures of specificity are already high; translated into clinical terms, they show that the application of the clinical marker early non-improvement is associated with a rate of false negative predictions (patients with non-improvement who nevertheless become responders or remitters with an unchanged treatment) of 2–19% for response and 0–13% for remission. This means that a medication change after two weeks in the individual patient with non-improvement would be inappropriate in 0–19% of patients, which is not desired and should be diminished to ideally 0%. The high consistency of the relation between early improvement/ favourable treatment outcome and early non-improvement/final treatment failure across several studies resulted in the postulation of a new paradigm for treatment response (Leuchter et al., 2009). The virtually identical time course of symptom amelioration after early improvement in patients treated with antidepressants of all classes or with placebo strongly suggests that early improvement and the successive time course of response reflect a common biological mechanism, which is not specific for a particular antidepressant medication. However, the neurobiological substrates of this remarkably robust relation between early improvement and final treatment outcome are not known and need to be elucidated. This is important, because it might promote approaches to increase remission rates by guiding antidepressant treatment by the use of biochemical markers for treatment response (Holsboer, 2008), either additional to the clinical marker early non-improvement or earlier in the course of an antidepressant treatment, e.g. after a few days of treatment. Substantial evidence from human and animal studies supports the involvement of neurotrophic factors in the pathophysiology of MDD (for extensive reviews, see Duman and Monteggia, 2006; Krishnan and Nestler, 2008). Studies in this field mainly focused on the brain-derived neurotrophic factor (BDNF), which is abundantly expressed in adult limbic structures. In the serum, the mean BDNF concentration in acutely depressed patients is significantly lower compared to healthy controls (Sen et al., 2008). The same report included a meta-analysis of eight clinical studies investigating the mean sBDNF concentration before and after antidepressant treatment, showing an increase of mean sBDNF concentration during antidepressant treatment, which achieved the BDNF level of the healthy control group. A second meta-analysis (Brunoni et al., 2008) confirmed these results and found a significant correlation between BDNF concentration and change in depression severity. Taken together, these analyses suggest that the increase of sBDNF is associated with a favourable treatment outcome and that the non-increase of sBDNF is associated with treatment failure in patients with MDD. The current study was performed with the main objective to investigate the potential clinical value of the marker “early nonincrease of sBDNF” for the guidance of treatment in patients with MDD. For this purpose, we investigated the following questions in inpatients with MDD, who were treated by antidepressant pharmacotherapy in a naturalistic setting: • Is there an association between an early sBDNF non-increase and an early HAMD-21 non-improvement? • Does the marker early non-increase of sBDNF (defined as nonincrease of sBDNF between baseline and day 7 or day 14, respectively) predict final non-response (defined as b50% sum score reduction in HAMD-21) and non-remission (defined as HAMD-21 sum score N7) (Frank et al., 1991)?
• Does the combined marker of early sBDNF non-increase plus early HAMD-21 non-improvement (defined as b20% sum score reduction in HAMD-21 between baseline and day 7 or day 14, respectively) predict final non-response or non-remission to a higher degree than the single markers early HAMD-21 non-improvement and early sBDNF non-increase? For each question we calculated specificity, sensitivity, positive and negative predictive value as markers of the predictive value according to established methods (Szegedi et al., 2003, 2009; Tadić et al., 2010a). 2. Methods 2.1. Patients Men and women, subsequently hospitalised at the Department of Psychiatry and Psychotherapy at the University Medical Centre Mainz for the treatment of MDD, participated in the present study. All patients gave their written informed consent after a complete description of the study. The study was approved by the local ethics committee of the Landesärztekammer Rheinland-Pfalz (study code n°: 837.405.08/6402) and is compliant with the Code of Ethics of the World Medical Association (Declaration of Helsinki). The design of the study was guided by the principle that results should be representative for inpatients with MDD and daily clinical practice; therefore, we used broad in-/exlusion criteria. In detail, inclusion criteria were i) Major Depressive Episode according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, American Psychiatric Association, 2004); ii) age between 18 and 65 years and ≤60 years at the beginning of the first depressive episode; iii) no treatment with an antidepressant medication or insufficient treatment response to an eventually existing antidepressant pre-medication (treatment duration N14 days), which was determined during a thorough clinical interview at admission to the hospital by at least one specialist in psychiatry (AT and/or KL) and at least one psychiatric resident; thus, all patients were about to be commenced or changed an antidepressant pharmacotherapy at the time of inclusion; and iv) written informed consent to study participation. Exclusion criteria were i) lifetime diagnosis of dementia, schizophrenia, schizoaffective disorder, or bipolar disorder according to DSM-IV; ii) current diagnosis of alcohol dependency (DSM-IV) requiring acute detoxification; iii) depression due to organic factors including Parkinson's disease or Multiple Sclerosis; iv) pregnancy or breast-feeding; and v) cognitive impairment, which precludes a correct psychometric assessment. Diagnosis of MDD according to DSM-IV criteria was ascertained by at least one resident and one specialist in psychiatry (AT or KL). One of two research assistants (KFS or LB), who had been trained in several rater trainings prior to the start of the study, additionally applied the German Version of the MINI International Neuropsychiatric Interview (Sheehan and Lecrubier, 1998). The antidepressant medication during the study period consisted of escitalopram (10–20 mg/day), sertraline (50–150 mg/ day), fluoxetine (20 mg/day), venlafaxine (150–375 mg/day), duloxetine (90–120 mg/day), mirtazapine (30–45 mg/day), tranylcypromine (30 mg/day), amitriptyline (225 mg/day), clomipramine (150 mg/day), or trimipramine (100 mg/day). Depression severity was assessed by means of the 21-item Hamilton Depression Rating Scale (HAMD-21) from baseline to day 42 in weekly intervals, which was applied by one of two trained research assistants (KFS or LB). 2.2. Measurement of serum BDNF levels For serum sampling blood was obtained in a serum separator tube from the antecubital vein between 08.00 and 11.00 am at baseline, days (d) 7, 14, 21, 28, 35, and 42. After 30 min of clotting time, the
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
whole blood was centrifugated at 1000 ×g for 15 min to isolate the serum. The serum was collected and kept at −80 °C before assaying BDNF concentration with an enzyme-linked immunosorbent assay kit (R&D Systems, Wiesbaden, Germany). 100 μl assay diluent and 50 μl BDNF standard or 50 μl thawed sample were added. After incubation for two hours at room temperature, 100 μl of mouse monoclonal antibody against BDNF conjugated to horseradish peroxidase was added and incubated for one hour at room temperature. The plates were washed using an autowasher and the appropriate wash buffer. After the third wash, any remaining wash buffer was removed. 200 μl of substrate solution consisting of color reagents mixed in equal volumes was added to produce a color reaction and the plates were incubated for 30 min protected from light. The color reaction was stopped with 2 N sulfuric acid. The absorbance at 450 nm was measured with a microplate reader (Model Sunrise, TECAN, Germany) to determine BDNF concentration according to the standard curve. Wavelength correction was conducted on 540 nm. Intra- and interassay coefficients of variation (CV) in our sample were 3.7% and 8.7%, respectively. The serum probes of each patient (BL–EP) were analysed on one ELISA plate; therefore, patients with a sBDNF change from baseline to day 7 ≤ + 3.7% were classified as having no sBDNF increase. 2.3. Statistical analysis Due to the naturalistic setting of the study, there were changes in antidepressant treatment (i.e., a medication switch or the initiation of an antidepressant combination following monotherapy) during the study period in some patients. An analysis of the relation between early changes of sBDNF and final treatment outcome in a sample with medication changes during the study period has to recognize the risk to investigate the relation between an absence of an early change of sBDNF level under treatment A, with a final treatment outcome, which however has been reached after the initiation of treatment B. Therefore, data analysis was performed on those data, which were obtained from baseline (BL) until the last visit before a treatment change (=endpoint, EP). Patients with observed data at BL, day 7 and at least one data between days 21 and 42 were eligible for analysis. Statistical analyses were performed exclusively on observed data. Differences between groups in baseline clinical and demographic data were analysed by t-tests for independent variables or Chi2-tests, when required. For statistical analysis, we defined the following predictors: Ia) sBDNF non-increase from BL to d7; Ib) sBDNF nonincrease from BL to d14; IIa) early non-improvement from BL to d7, defined as a HAMD-21 sum score decrease b20% from BL to d7; IIb) early non-improvement from BL to d14, defined as a HAMD-21 sum score decrease b20% from BL to d14; IIIa) sBDNF non-increase from BL to d7 plus early non-improvement from BL to d7; IIIb) sBDNF nonincrease from BL to d14 plus early non-improvement from BL to d14. Outcome parameters were HAMD-21 non-response at endpoint, defined as HAMD sum score decrease b50%, and HAMD-21 nonremission, defined as HAMD-21 sum score N7 (Frank et al., 1991). Based on these data, we calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for the prediction of outcomes. Sensitivity in this context means the proportion of non-responders who are identified by early nonimprovement or sBDNF non-increase, respectively. Specificity in this context means the proportion of responders who are identified by improvement or increase. PPV means the rate of patients with early non-improvement who become non-responder or non-remitter/all non-improver ⁎ 100, and NPV means early improver who become responder or remitter/all improver ⁎ 100. Aditionally, we calculated Odds-ratios (ORs) and 95%-confidence intervals, providing a measure of effect size, describing the strength of the association and nonindependence between two binary data values. We performed ROC analyses to calculate the area under the curve (ROC-AUC) and the
417
confidence intervals around the ROC-AUC for the prediction of a sBDNF/HAMD-21 change from baseline to days 7 and 14 on response and remission at endpoint. Furthermore, we performed a ROC analyses to determine the most appropriate cut-off for the prediction of a sBDNF change from baseline to day 7 on treatment response at the endpoint. Significance was set at p ≤ 0.050. All analyses were done using Predictive Analysis SoftWare (PASW) 18.0. 3. Results 3.1. Baseline demographical and clinical characteristics as well as data of predictors and outcomes The ITT sample consisted of 46 patients; five patients had to be excluded from analysis because of missing data, resulting in 41 (49% men, 51% women) eligible subjects. Of those, 24% had the endpoint (EP) at day (d) 21, 32% at d28, 12% at d35 and 32% at d42. Mean (±SD) age was 44.8 ± 13.2 years (range: 20–64 years). Eighty percent suffered from a recurrent depressive disorder, 20% from their first depressive episode. Mean (±SD) HAMD-21 sum score at baseline was 20.8 ± 4.5 pts. (range: 10–31). Mean (±SD) sBDNF level at baseline was 22.5 ± 9.2 ng/ml (range: 7.0–51.4 ng/ml). Patients displayed “no sBDNF increase” from baseline to day 7 in 51%, and from baseline to day 14 in 34%. Patients were HAMD-21 non-improver on day 7 and day 14 in 55% and 43%, respectively. The combined marker non-increase of sBDNF plus non-improvement from baseline to day 7 or day 14 was present in 59% and 25%, respectively. 71% was non-responder at the endpoint, 81% was non-remitter. 3.2. Predictive value of early sBDNF non-increase for early HAMD-21 non-improvement The predictive values of a sBDNF non-increase from baseline to days 7 and 14 for HAMD improvement at days 7 and 14, respectively, are shown in Table 1. The sBDNF non-increase from BL to d7 predicted a HAMD-21 non-improvement from BL to d7 with moderate sensitivity and specificity; a sBDNF non-increase from BL to d14 predicted a HAMD-21 non-improvement from BL to d14 with low to moderate sensitivity and specificity. 3.3. Predictive value of early sBDNF non-increase for final non-response and non-remission The predictive values of the marker sBDNF non-increase from BL to days 7 and 14, respectively, for the treatment outcomes non-response and non-remission are shown in Table 2. A sBDNF non-increase from BL to days 7 and 14 predicted final non-response with 67% and 73% specificity, resp., and final non-remission with 63% and 71%, resp. specificity. Sensitivity for later non-response and non-remission was markedly lower. NPV for non-response (40 and 32%, resp.) and nonremission (25 and 20%, resp.) were far lower than the PPV (81 and 79%, resp.; 86 and 86%, resp). The area under the curve (ROC-AUC) for the prediction of a sBDNF change from baseline to days 7 and 14 on non-response at the endpoint was 60.3% (CI: 39.8–80.8%) and 59.7% (CI: 40.6–78.8%), respectively. The ROC-AUC for the prediction of a sBDNF change from baseline to days 7 and 14 on non-remission at the endpoint was 60.4% (CI: 35.6–84.2%) and 58.5% (CI: 37.1–79.9%), respectively. 3.4. Predictive value of early HAMD-21 non-improvement for final nonresponse and non-remission The predictive values of HAMD-21 non-improvement from BL to days 7 and 14, respectively, for the treatment outcomes non-response and non-remission are shown in Table 3. A HAMD-21 nonimprovement from baseline to days 7 and 14 was moderately to
418
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
Table 1 Predictive value of sBDNF non-increase from baseline to day 7 or 14 on HAMD-21 improvement on day 7 or 14, respectively. Predictor
Outcome (N) Non-Improvement d7
Improvement d7
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
57% (40.6–72.3)
56% (39.7–71.4)
62% (45.5–76.7)
50% (34.0–66.0)
1.63 (0.0–11.5)
44%
24% (11.8–40.3)
55% (38.3–70.9)
29% (15.6–45.7)
48% (31.8–64.6)
0.37 (0.0–9.7)
45%
sBDNF d7
Non-increase Increase
13 10
8 10
sBDNF d14a
Non-increase Increase
Non-Improvement d14 4 13
Improvement d14 10 12
CI: confidence interval; d: day; FDR: false discovery rate; non-improvement d7 (d14): HAMD-21 sum score decrease b20% between baseline and day 7 (14); NPV: negative predictive value. PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a N = 39 due to missing data of two patients at d14.
highly specific for later non-response (64 and 91%, resp.) and nonremission (57 and 86%, resp.). Sensitivity for later non-response and non-remission was lower. NPV for non-response and non-remission was far lower than the PPV (81 and 94%, resp.; 86 and 94%, resp). The ROC-AUC for the prediction of a HAMD-21 change from baseline to days 7 and 14 on non-response at the endpoint was 66.5% (CI: 44.5– 88.5%) and 85.4% (CI: 72.6–98.3%), respectively; the ROC-AUC for the prediction of a HAMD-21 change from baseline to days 7 and 14 on non-remission at the endpoint was 62.9% (CI: 35.2–90.6%) and 87.4% (CI: 00.0–100.0%), respectively.
3.5. Predictive value of an early sBDNF non-increase plus early HAMD-21 non-improvement for final non-response and non-remission The predictive values of the combined marker sBDNF non-increase plus HAMD-21 non-improvement from baseline to days 7 and 14, respectively, for the treatment outcomes non-response and nonremission are shown in Table 4. A sBDNF non-increase and HAMD-21 non-improvement from BL to days 7 and 14 were highly specific for later non-response (67 and 100%, resp.) and non-remission (60 and 100%, resp.). Sensitivity for later non-response and non-remission was lower. NPV for non-response and non-remission was far lower than the PPV (67 and 58%, resp.; 77 and 100%, resp.).
3.6. Determination of the most appropriate cut-off value for the prediction of treatment response at the endpoint by early sBDNF change ROC analysis revealed a sBDNF change of ± 4.5% between baseline and day 7 as optimal cut-off for the prediction of treatment response at EP; this value was nearly identical to the pre-defined cut-off of ± 3.7%.
4. Discussion Our pilot study provides evidence that the individual marker early sBDNF non-increase has a positive predictive value for final treatment failure very similar to the clinical marker early nonimprovement. This was not due to the fact that the non-increase of sBDNF is the major underlying biological factor of early-nonimprovement as documented by the additional analysis showing only a small overlap between early BDNF non-increase and early non-improvement. The non-increase of sBDNF has only a limited predictive value for early non-improvement, suggesting that sBDNF increase is only one of many other factors underlying the provocation of early improvement by antidepressant treatment. The biological reasons for the similar predictive values on treatment outcome are therefore unclear. It might be that the sBDNF change and the biological changes underlying early improvement are overlapping and lead to a common pathway underlying the course of symptom amelioration. In our study, the combined marker sBDNF non-increase plus early non-improvement has a higher predictive value compared to the individual markers sBDNF non-increase and early non-improvement: at week 2, the combined marker reaches a positive predictive value for treatment failure of 100%, thereby decreasing the rate of false negative predictions to 0%.Translated into clinical terms this means, the clinician could change the treatment strategy after two weeks of treatment, when he is observing a non-increase of sBDNF and a HAMD-21 non-improvement from treatment initiation to day 14, because no patient with these characteristics will experience a subsequent response or remission. In biological terms, the increase of the predictive value of the combined marker compared to the single markers again suggests different but overlapping biologicial entities of sBDNF increase on the one hand and the biology of early improvement on the other hand.
Table 2 Predictive value of sBDNF non-increase from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Predictor
sBDNF d7 sBDNF d14⁎
sBDNF d7 sBDNF d14a
Outcome (N) Non-Response
Response
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
Non-increase Increase Non-increase Increase
17 12 11 17
4 8 3 8
59% (42.6–74.1)
67% (50.6–80.9)
81% (65.7–91.5)
40% (25.0–56.5)
2.83 (0.1–13.5)
33%
39% (23.8–55.9)
73% (56.4–85.9)
79% (63.0–90.4)
32% (18.0–48.8)
1.73 (0.0–12.1)
25%
Non-increase Increase Non-increase Increase
Non-Remission 18 15 12 20
Remission 3 5 2 5
55% (38.7–70.6)
63% (46.5–77.5)
86% (71.6–94.8)
25% (12.8–41.0)
2.00 (0.0–12.1)
38%
38% (23.0–54.9)
71% (54.3–84.4)
86% (71.1–95.0)
20% (8.9–35.9)
1.50 (0.0–11.7)
28%
CI: Confidence interval; d: day; FDR: false discovery rate; non-response: HAMD-21 sum score decrease b50% between baseline and endpoint; non-remission: HAMD-21 sum score N 7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a N = 39 due to missing data of two patients at d14.
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
419
Table 3 Predictive value of HAMD non-improvement from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Predictor
HAMD d7 HAMD d14⁎
HAMD d7 HAMD d14a
Outcome (N) Non-Response
Response
Sensitivity (95% CL)
Specificity (95% CL)
PPV (95% CL)
NPV (95% CL)
Odds-Ratio (95% CL)
FDR
Non-improvement Improvment Non-improvement Improvment
19 10 16 13
4 8 1 10
66% (49.6–80.0)
67% (50.6–80.9)
83% (68.0–92.9)
44% (28.6–60.3)
3.80 (0.3–15.0)
33%
55% (38.5–70.7)
91% (77.6–97.7)
94% (81.7–99.0)
43% (27.5–59.6)
12.31 (4.0–26.8)
9%
Non-improvement Improvment Non-improvement Improvment
Non-Remission 20 13 16 17
Remission 3 5 1 6
61% (44.5–75.8)
63% (46.5–77.5)
87% (72.8–95.4)
28% (15.1–44.2)
1.85 (0.1–13.0)
38%
48% (32.0–64.3)
86% (71.4–94.9)
94% (81.7–99.0)
26% (13.4–42.3)
5.65 (0.8–17.8)
14%
CI: confidence interval; d: day; FDR: false discovery rate; Non-improvement d7 (d14): HAMD-21 sum score decrease b 20% between baseline and day 7 (14); Non-Response: HAMD21 sum score decrease b50% between baseline and endpoint; Non-Remission: HAMD-21 sum score N7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value. a N = 39 due to missing data of two patients at d14.
The results of this study suggest a valuable clinical utility of peripheral BDNF measurement as surrogate marker of antidepressant efficacy, as it has been proposed earlier (Brunoni et al., 2008; Sen et al., 2008; MacQueen, 2009). They provide further evidence for the importance of early changes of sBDNF in antidepressant treatment outcome and extend the existing results in both animals and humans: one study in humans investigated sBDNF levels at baseline and shortly after the initiation of antidepressant treatment (Gorgulu and Caliyurt, 2009). In this study, patients with MDD showed an increase of sBNDF within one day after sleep deprivation. Rats, which were treated with various antidepressant substances, showed a BDNF increase by posttranslational mechanisms within seven days in the hippocampus and the prefrontal-/frontal cortex (Musazzi et al., 2009). Taken together, the results support the hypothesis that sBDNF can be increased rapidly by the initiation of antidepressant treatment and – in turn – the absence of an early increase of sBDNF seems to indicate that the selected treatment will not be effective. It is not clear as to what extent peripheral BDNF reflects central BDNF (Sen et al., 2008; MacQueen, 2009). However, given that antidepressant treatment increases central BDNF expression (Nibuya et al., 1995; Malberg et al., 2000) and BDNF increase is associated with depressive symptom amelioration, the results in our study support
the idea of a close relation between peripheral and central BDNF. The explanation might be the existence of a bi-directional transport of BDNF crossing the blood-brain barrier, consisting of an active transport system for BDNF for the passage from blood to brain and the reabsorption from central spinal fluid to blood (Pan et al., 1998). One advantage of our pilot study is the repeated measurement of sBDNF levels starting shortly after the initiation of antidepressant treatment, because it enabled the investigation of the surrogate marker “sBDNF non-increase” in a time span of 7 and 14 days after treatment initiation. This could represent a progress in clinical decision making for the individual patient compared to the current practice of waiting for improvement during 4–6 weeks after treatment initiation. The results of our study with an association of sBDNF non-increase and treatment failure in numerous but not all individual patients of this study also show that BDNF is not able to fully encompass the complexity of MD. The close-meshed measurement of a biomarker of interest and the combination of biochemical and clinical markers might be useful for several other biological target systems presumably involved in the treatment response of major depression. The binarization of the sBDNF change from BL to days 7/14 as a predictor is a simplification which might have led to a loss of information, but mirrors what could be feasible and applicable in daily clinical practice. The herein defined
Table 4 Predictive value of sBDNF non-increase and HAMD-21 non-improvement from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Outcome (N)a
Predictor
Non-Response sBDNF + HAMD d7
sBDNF + HAMD d14
sBDNF + HAMD d7
sBDNF + HAMD d14
Non-increase + non-improvement Increase + improvement Non-increase + non-improvement Increase + improvement
Non-increase + non-improvement Increase + improvement Non-increase + non-improvement Increase + improvement
Response
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
3
77% (54.3–92.0)
67% (43.9–85.3)
77% (54.3–92.0)
67% (37.7–79.6)
6.67 (0.5–25.9)
33%
44% (19.9–70.3)
100% (79.4–100)
100% (79.4–100)
58% (31.4–81.6)
0.00
65% (42.0–83.8)
60% (37.2–80.0)
85% (63.4–96.5)
33% (14.7–56.1)
2.75 (0.0–20.1)
33% (12.1–60.4)
100% (79.4–100)
100% (79.4–100)
33% (12.1–60.4)
0.00
6 4
0
5
7
Non-Remission 11
Remission 2
6
3
4
0
8
4
0%
40%
0%
CI: confidence interval; d: day; FDR: false discovery rate; non-improvement d7 (d14): HAMD-21 sum score decrease b20% between baseline and day 7 (14); non-response: HAMD21 sum score decrease b50% between baseline and endpoint; non-remission: HAMD-21 sum score N7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a Reduced N, because analysis included extreme groups, i.e. subjects with sBDNF non-increase plus HAMD non-improvement as well as sBDNF increase plus HAMD improvement; subjects with sBDNF non-increase plus HAMD improvement and sBDNF increase plus HAMD non-improvement were not incorporated in the analysis.
420
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
cut-off has to be validated in future studies with healthy controls, larger samples of MD patients as well as in patients with different psychiatric disorders. Finally, the size of our sample is small and was associated with a false discovery rate between 0% and 45%; therefore, the results of our pilot study have to be interpreted cautiously and need validation in independent samples. 5. Conclusion In conclusion, the results of our study are promising in regard to the potential identification of a biomarker for treatment failure in patients with MDD which can be easily assessed in clinical routine. This is very important, because a biomarker for treatment nonresponse in MDD is warranted for both clinical application and industrial research. An independent replication of the results showing that the combined marker early-non-increase plus early-non-improvement predicts the final treatment failure to 100% could represent the basis for randomized controlled trials investigating its clinical value for the guidance of antidepressant treatment. Furthermore, the results of our study might provide a valuable model for the investigation of biomarkers in the treatment of MDD and other psychiatric disorders and might encourage research in this area in order to further shorten the duration of psychopharmacological treatment until the determination of insufficient effectiveness. Finally, these studies should lead to clear decision plans for guidance of psychopharmacological treatment of the individual patient. Financial disclosures None of the authors reported biomedical financial interests or potential conflicts of interest. Acknowledgements We are grateful to all participants for the support of study as well as to Dr. Tanja Zeller for the technical assistance. References American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders4th ed. . Washington DC: American Psychiatric Press; 2004. Brunoni AR, Lopes M, Fregni F. A systematic review and meta-analysis of clinical studies on major depression and BDNF levels: implications for the role of neuroplasticity in depression. Int J Neuropsychopharmacol 2008;11:1169–80. Duman RS, Monteggia LM. A neurotrophic model for stress-related mood disorders. Biol Psychiatry 2006;59:1116–27. Fournier JC, DeRubeis RJ, Hollon SD, Dimidjian S, Amsterdam JD, Shelton RC, et al. Antidepressant drug effects and depression severity: a patient-level meta-analysis. JAMA 2010;303:47–53. Frank E, Prien RF, Jarrett RB, Keller MB, Kupfer DJ, Lavori PW, et al. Conceptualization and rationale consensus definitions of terms in major depressive disorder. Remission, recovery and recurrence. Arch Gen Psychiatry 1991;48:851–5.
Gorgulu Y, Caliyurt O. Rapid antidepressant effects of sleep deprivation therapy correlates with serum BDNF changes in major depression. Brain Res Bull 2009;80:158–62. Henkel V, Seemüller F, Obermeier M, Adli M, Bauer M, Mundt C, et al. Does early improvement triggered by antidepressants predict response/remission? Analysis of data from a naturalistic study on a large sample of inpatients with major depression. J Affect Dis 2009;115:439–49. Hennings JM, Owashi T, Binder EB, Horstmann S, Menke A, Kloiber S, et al. Clinical characteristics and treatment outcome in a representative sample of depressed inpatients. Findings from the Munich Antidepressant Response Signature (MARS) project. Psychiatric Res 2009;43:215–29. Holsboer F. How can we realize the promise of personalized antidepressant medicines? Nat Rev Neurosci 2008;9:638–46. Katz MM, Tekell JL, Bowden CL, Brannan S, Houston JP, Berman N, et al. Onset and early behavioral effects of pharmacologically different antidepressants and placebo in depression. Neuropsychopharmacology 2004;29:566–79. Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ, Johnson BT. Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. Public Libr Sci Med 2008;5:e45. Krishnan V, Nestler EJ. The molecular neurobiology of depression. Nature 2008;455: 894–902. Leuchter AF, Cook IA, Hunter AM, Korb AS. A new paradigm for the prediction of antidepressant treatment response. Dialogues Clin Neurosci 2009;11:435–46. MacQueen GM. Magnetic resonance imaging and prediction of outcome in patients with major depressive disorder. J Psychiatry Neurosci 2009;34:343–9. Malberg JE, Eisch AJ, Nestler EJ, Duman RS. Chronic antidepressant treatment increases neurogenesis in adult rat hippocampus. J Neurosci 2000;20:9104–10. Musazzi L, Cattaneo A, Tardito D, Barbon A, Gennarelli M, Barlati S, et al. Early raise of BDNF in hippocampus suggests induction of posttranscriptional mechanisms by antidepressants. Biomed Cent Neurosci 2009;10:48. Nibuya M, Morinobu S, Duman RS. Regulation of BDNF and trkB mRNA in rat brain by chronic electroconvulsive seizure and antidepressant drug treatments. J Neurosci 1995;15:7539–47. Nierenberg AA, Farabaugh AH, Alpert JE, Gordon J, Worthington JJ, Rosenbaum JF, et al. Timing of onset of antidepressant response with fluoxetine treatment. Am J Psychiatry 2000;157:1423–8. Pan W, Banks WA, Fasold MB, Bluth J, Kastin AJ. Transport of brain-derived neurotrophic factor across the blood-brain barrier. Neuropharmacology 1998;37:1553–61. Sen S, Duman R, Sanacora G. Serum brain-derived neurotrophic factor, depression, and antidepressant medications: meta-analyses and implications. Biol Psychiatry 2008;64:527–32. Sheehan DV, Lecrubier Y. M.I.N.I. International Neuropsychiatric Interview. USA Tampa; 1998. Stassen HH, Angst J, Hell D, Scharfetter C, Szegedi A. Is there a common resilience mechanism underlying antidepressant drug response? Evidence from 2848 patients. J Clin Psychiatry 2007;68:1195–205. Szegedi A, Müller MJ, Anghelescu I, Klawe C, Kohnen R, Benkert O. Early improvement under mirtazapine and paroxetine predicts later stable response and remission with high sensitivity in patients with major depression. J Clin Psychiatry 2003;64:413–20. Szegedi A, Jansen WT, van Willigenburg AP, van der Meulen E, Stassen HH, Thase ME. Early improvement as a predictor of treatment outcome in patients with major depressive disorder: why the first 2 weeks really matter-evidence from 6,562 patients. J Clin Psychiatry 2009;70:344–53. Tadić A, Helmreich I, Mergl R, Hautzinger M, Henkel V, Hegerl U. Early improvement is a predictor of treatment outcome in patients with mild major, minor or subsyndromal depression. J Affect Dis 2010a;120:86–93. Tadić A, Gorbulev S, Dahmen N, Hiemke C, Braus DF, Röschke J, et al. EMC Study Group. Rationale and design of the randomised clinical trial comparing early medication change (EMC) strategy with treatment as usual (TAU) in patients with major depressive disorder—the EMC trial. Trials 2010b;11:21. Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R. Selective publication of antidepressant trials and its influence on apparent efficacy. N Engl J Med 2008;358: 252–60. Warden D, Rush AJ, Trivedi MH, Fava M, Wisniewski SR. The STAR*D Project results: a comprehensive review of findings. Curr Psychiatry Rep 2007;9:449–59.
Contents lists available at ScienceDirect
Progress in Neuro-Psychopharmacology & Biological Psychiatry j o u r n a l h o m e p a g e : w w w. e l s ev i e r. c o m / l o c a t e / p n p
The early non-increase of serum BDNF predicts failure of antidepressant treatment in patients with major depression: A pilot study André Tadić a,⁎,1, Stefanie Wagner a,1, Konrad Friedrich Schlicht a, Dirk Peetz b, Liudmyla Borysenko a, Nadine Dreimüller a, Christoph Hiemke a, Klaus Lieb a a b
Department of Psychiatry and Psychotherapy, University Medical Centre Mainz, Germany Institute for Clinical Chemistry and Laboratory Medicine, University Medical Centre Mainz, Germany
a r t i c l e
i n f o
Article history: Received 28 June 2010 Received in revised form 9 August 2010 Accepted 15 August 2010 Available online 20 August 2010 Keywords: Biomarker Brain-derived neurotrophic factor Major depressive disorder Serum Treatment outcome
a b s t r a c t In the treatment of patients with major depressive disorder (MDD), early non-improvement of symptoms after initiation of antidepressant treatment is a highly sensitive and specific marker for final treatment failure. On the other hand, meta-analyses of clinical studies investigating serum BDNF (sBDNF) concentration before and after antidepressant treatment showed an increase of sBDNF during treatment, which was correlated with amelioration of depressive symptoms. No study has yet investigated the predictive value of early changes of sBDNF for final treatment outcome of the individual patient. The aim of this study was to investigate in patients with MDD, whether i) the non-increase of sBDNF in the early course of treatment is a specific and sensitive marker for final treatment failure, ii) whether the sensitivity and specificity of early non-improvement for treatment failure can be increased by combining it with the marker “early nonincrease of sBDNF”. For this purpose, we performed a pilot study with 41 inpatients with MDD according to DSM-IV, who were treated in a naturalistic setting. Depression severity and sBDNF were measured in weekly intervals from baseline to week six with the 21-item Hamilton Depression Rating Scale (HAMD-21) and ELISA, respectively. The individual markers sBDNF non-increase and HAMD-21 non-improvement from baseline to day 7 or 14 predicted later non-response and non-remission with moderate to high specificity. The combined marker sBDNF non-increase plus HAMD-21 non-improvement at day 14 increased the specificity for non-response and non-remission to 100%. Our data provide the first evidence that the absence of an early increase of sBDNF in conjunction with early non-improvement might be a highly specific peripheral marker predictive for treatment failure in patients with MDD. If replicated, this combined marker could be considered useful for prospective confirmatory trials in patients with MDD. © 2010 Elsevier Inc. All rights reserved.
1. Introduction In the treatment of Major Depressive Disorder (MDD) outcomes with currently available antidepressants are disappointing, because only 30–40% of “real-life” patients with MDD experience a remission Abbreviations: AD, antidepressant; BL, baseline; D7, day 7; D14, day 14; DSM-IV, Diagnostic and Statistical Manual of Mental Disorders; ELISA, enzyme-linked immunosorbent assay; EP, endpoint; HAMD-21, 21-item version of the Hamilton Depression Rating Scale; ITT, intention-to-treat; MDD, major depressive disorder; NPV, negative predictive value; PASW, Predictive Analysis SoftWare; PPV, positive predictive value; sBDNF, serum brain-derived neurotrophic factor; SD, standard deviation. ⁎ Corresponding author. Department of Psychiatry and Psychotherapy, University Medical Centre of the Johannes-Gutenberg-University Mainz, Untere Zahlbacher Str. 8, Germany. Tel.: + 49 6131 17 3950; fax: + 49 6131 17 3459. E-mail addresses: [email protected] (A. Tadić), [email protected] (S. Wagner), [email protected] (K.F. Schlicht), [email protected] (D. Peetz), [email protected] (L. Borysenko), [email protected] (N. Dreimüller), [email protected] (C. Hiemke), [email protected] (K. Lieb). 1 These authors contributed equally to this manuscript. 0278-5846/$ – see front matter © 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.pnpbp.2010.08.011
with the first selected antidepressant (Warden et al., 2007, and refs. inside). Their effect sizes are rather small than medium (Kirsch et al., 2008; Turner et al., 2008; Fournier et al., 2010) and more effective antidepressant substances cannot be expected in the near future. Therefore, it is indispensable to develop new strategies to increase remission rates and to shorten the time to remission in acutely depressed patients. One approach is the identification of individual patients who will not respond to a selected antidepressant treatment as early as possible and a consecutive treatment change in these patients. A substantial body of evidence from many retrospective studies incorporating in total more than 12,000 patients and virtually all groups of antidepressants (Henkel et al., 2009; Hennings et al., 2009; Katz et al., 2004; Nierenberg et al., 2000; Stassen et al., 2007; Szegedi et al., 2003, 2009; Tadić et al., 2010a) shows that non-improvement, oftenly defined as b20% reduction of depressive symptoms (assessed with a routine scale like the Hamilton Depression Rating Scale (HAMD)) after 14 days of treatment is a highly specific marker for final treatment failure with an unchanged treatment. E.g., the largest
416
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
meta-analysis in this field (Szegedi et al., 2009) showed that nonimprovement after two weeks of treatment is a highly specific predictor for final non-response (81–98% across the investigated studies) and non-remission (87–100%). Based on these data, the application of the marker early non-improvement in clinical routine might be able to reduce the treatment time with an insufficient drug to two weeks. Its potential to increase remission rates in the treatment of acutely depressed patients is currently investigated by a large randomized clinical trial (Tadić et al., 2010b). The above mentioned figures of specificity are already high; translated into clinical terms, they show that the application of the clinical marker early non-improvement is associated with a rate of false negative predictions (patients with non-improvement who nevertheless become responders or remitters with an unchanged treatment) of 2–19% for response and 0–13% for remission. This means that a medication change after two weeks in the individual patient with non-improvement would be inappropriate in 0–19% of patients, which is not desired and should be diminished to ideally 0%. The high consistency of the relation between early improvement/ favourable treatment outcome and early non-improvement/final treatment failure across several studies resulted in the postulation of a new paradigm for treatment response (Leuchter et al., 2009). The virtually identical time course of symptom amelioration after early improvement in patients treated with antidepressants of all classes or with placebo strongly suggests that early improvement and the successive time course of response reflect a common biological mechanism, which is not specific for a particular antidepressant medication. However, the neurobiological substrates of this remarkably robust relation between early improvement and final treatment outcome are not known and need to be elucidated. This is important, because it might promote approaches to increase remission rates by guiding antidepressant treatment by the use of biochemical markers for treatment response (Holsboer, 2008), either additional to the clinical marker early non-improvement or earlier in the course of an antidepressant treatment, e.g. after a few days of treatment. Substantial evidence from human and animal studies supports the involvement of neurotrophic factors in the pathophysiology of MDD (for extensive reviews, see Duman and Monteggia, 2006; Krishnan and Nestler, 2008). Studies in this field mainly focused on the brain-derived neurotrophic factor (BDNF), which is abundantly expressed in adult limbic structures. In the serum, the mean BDNF concentration in acutely depressed patients is significantly lower compared to healthy controls (Sen et al., 2008). The same report included a meta-analysis of eight clinical studies investigating the mean sBDNF concentration before and after antidepressant treatment, showing an increase of mean sBDNF concentration during antidepressant treatment, which achieved the BDNF level of the healthy control group. A second meta-analysis (Brunoni et al., 2008) confirmed these results and found a significant correlation between BDNF concentration and change in depression severity. Taken together, these analyses suggest that the increase of sBDNF is associated with a favourable treatment outcome and that the non-increase of sBDNF is associated with treatment failure in patients with MDD. The current study was performed with the main objective to investigate the potential clinical value of the marker “early nonincrease of sBDNF” for the guidance of treatment in patients with MDD. For this purpose, we investigated the following questions in inpatients with MDD, who were treated by antidepressant pharmacotherapy in a naturalistic setting: • Is there an association between an early sBDNF non-increase and an early HAMD-21 non-improvement? • Does the marker early non-increase of sBDNF (defined as nonincrease of sBDNF between baseline and day 7 or day 14, respectively) predict final non-response (defined as b50% sum score reduction in HAMD-21) and non-remission (defined as HAMD-21 sum score N7) (Frank et al., 1991)?
• Does the combined marker of early sBDNF non-increase plus early HAMD-21 non-improvement (defined as b20% sum score reduction in HAMD-21 between baseline and day 7 or day 14, respectively) predict final non-response or non-remission to a higher degree than the single markers early HAMD-21 non-improvement and early sBDNF non-increase? For each question we calculated specificity, sensitivity, positive and negative predictive value as markers of the predictive value according to established methods (Szegedi et al., 2003, 2009; Tadić et al., 2010a). 2. Methods 2.1. Patients Men and women, subsequently hospitalised at the Department of Psychiatry and Psychotherapy at the University Medical Centre Mainz for the treatment of MDD, participated in the present study. All patients gave their written informed consent after a complete description of the study. The study was approved by the local ethics committee of the Landesärztekammer Rheinland-Pfalz (study code n°: 837.405.08/6402) and is compliant with the Code of Ethics of the World Medical Association (Declaration of Helsinki). The design of the study was guided by the principle that results should be representative for inpatients with MDD and daily clinical practice; therefore, we used broad in-/exlusion criteria. In detail, inclusion criteria were i) Major Depressive Episode according to the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV, American Psychiatric Association, 2004); ii) age between 18 and 65 years and ≤60 years at the beginning of the first depressive episode; iii) no treatment with an antidepressant medication or insufficient treatment response to an eventually existing antidepressant pre-medication (treatment duration N14 days), which was determined during a thorough clinical interview at admission to the hospital by at least one specialist in psychiatry (AT and/or KL) and at least one psychiatric resident; thus, all patients were about to be commenced or changed an antidepressant pharmacotherapy at the time of inclusion; and iv) written informed consent to study participation. Exclusion criteria were i) lifetime diagnosis of dementia, schizophrenia, schizoaffective disorder, or bipolar disorder according to DSM-IV; ii) current diagnosis of alcohol dependency (DSM-IV) requiring acute detoxification; iii) depression due to organic factors including Parkinson's disease or Multiple Sclerosis; iv) pregnancy or breast-feeding; and v) cognitive impairment, which precludes a correct psychometric assessment. Diagnosis of MDD according to DSM-IV criteria was ascertained by at least one resident and one specialist in psychiatry (AT or KL). One of two research assistants (KFS or LB), who had been trained in several rater trainings prior to the start of the study, additionally applied the German Version of the MINI International Neuropsychiatric Interview (Sheehan and Lecrubier, 1998). The antidepressant medication during the study period consisted of escitalopram (10–20 mg/day), sertraline (50–150 mg/ day), fluoxetine (20 mg/day), venlafaxine (150–375 mg/day), duloxetine (90–120 mg/day), mirtazapine (30–45 mg/day), tranylcypromine (30 mg/day), amitriptyline (225 mg/day), clomipramine (150 mg/day), or trimipramine (100 mg/day). Depression severity was assessed by means of the 21-item Hamilton Depression Rating Scale (HAMD-21) from baseline to day 42 in weekly intervals, which was applied by one of two trained research assistants (KFS or LB). 2.2. Measurement of serum BDNF levels For serum sampling blood was obtained in a serum separator tube from the antecubital vein between 08.00 and 11.00 am at baseline, days (d) 7, 14, 21, 28, 35, and 42. After 30 min of clotting time, the
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
whole blood was centrifugated at 1000 ×g for 15 min to isolate the serum. The serum was collected and kept at −80 °C before assaying BDNF concentration with an enzyme-linked immunosorbent assay kit (R&D Systems, Wiesbaden, Germany). 100 μl assay diluent and 50 μl BDNF standard or 50 μl thawed sample were added. After incubation for two hours at room temperature, 100 μl of mouse monoclonal antibody against BDNF conjugated to horseradish peroxidase was added and incubated for one hour at room temperature. The plates were washed using an autowasher and the appropriate wash buffer. After the third wash, any remaining wash buffer was removed. 200 μl of substrate solution consisting of color reagents mixed in equal volumes was added to produce a color reaction and the plates were incubated for 30 min protected from light. The color reaction was stopped with 2 N sulfuric acid. The absorbance at 450 nm was measured with a microplate reader (Model Sunrise, TECAN, Germany) to determine BDNF concentration according to the standard curve. Wavelength correction was conducted on 540 nm. Intra- and interassay coefficients of variation (CV) in our sample were 3.7% and 8.7%, respectively. The serum probes of each patient (BL–EP) were analysed on one ELISA plate; therefore, patients with a sBDNF change from baseline to day 7 ≤ + 3.7% were classified as having no sBDNF increase. 2.3. Statistical analysis Due to the naturalistic setting of the study, there were changes in antidepressant treatment (i.e., a medication switch or the initiation of an antidepressant combination following monotherapy) during the study period in some patients. An analysis of the relation between early changes of sBDNF and final treatment outcome in a sample with medication changes during the study period has to recognize the risk to investigate the relation between an absence of an early change of sBDNF level under treatment A, with a final treatment outcome, which however has been reached after the initiation of treatment B. Therefore, data analysis was performed on those data, which were obtained from baseline (BL) until the last visit before a treatment change (=endpoint, EP). Patients with observed data at BL, day 7 and at least one data between days 21 and 42 were eligible for analysis. Statistical analyses were performed exclusively on observed data. Differences between groups in baseline clinical and demographic data were analysed by t-tests for independent variables or Chi2-tests, when required. For statistical analysis, we defined the following predictors: Ia) sBDNF non-increase from BL to d7; Ib) sBDNF nonincrease from BL to d14; IIa) early non-improvement from BL to d7, defined as a HAMD-21 sum score decrease b20% from BL to d7; IIb) early non-improvement from BL to d14, defined as a HAMD-21 sum score decrease b20% from BL to d14; IIIa) sBDNF non-increase from BL to d7 plus early non-improvement from BL to d7; IIIb) sBDNF nonincrease from BL to d14 plus early non-improvement from BL to d14. Outcome parameters were HAMD-21 non-response at endpoint, defined as HAMD sum score decrease b50%, and HAMD-21 nonremission, defined as HAMD-21 sum score N7 (Frank et al., 1991). Based on these data, we calculated sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) for the prediction of outcomes. Sensitivity in this context means the proportion of non-responders who are identified by early nonimprovement or sBDNF non-increase, respectively. Specificity in this context means the proportion of responders who are identified by improvement or increase. PPV means the rate of patients with early non-improvement who become non-responder or non-remitter/all non-improver ⁎ 100, and NPV means early improver who become responder or remitter/all improver ⁎ 100. Aditionally, we calculated Odds-ratios (ORs) and 95%-confidence intervals, providing a measure of effect size, describing the strength of the association and nonindependence between two binary data values. We performed ROC analyses to calculate the area under the curve (ROC-AUC) and the
417
confidence intervals around the ROC-AUC for the prediction of a sBDNF/HAMD-21 change from baseline to days 7 and 14 on response and remission at endpoint. Furthermore, we performed a ROC analyses to determine the most appropriate cut-off for the prediction of a sBDNF change from baseline to day 7 on treatment response at the endpoint. Significance was set at p ≤ 0.050. All analyses were done using Predictive Analysis SoftWare (PASW) 18.0. 3. Results 3.1. Baseline demographical and clinical characteristics as well as data of predictors and outcomes The ITT sample consisted of 46 patients; five patients had to be excluded from analysis because of missing data, resulting in 41 (49% men, 51% women) eligible subjects. Of those, 24% had the endpoint (EP) at day (d) 21, 32% at d28, 12% at d35 and 32% at d42. Mean (±SD) age was 44.8 ± 13.2 years (range: 20–64 years). Eighty percent suffered from a recurrent depressive disorder, 20% from their first depressive episode. Mean (±SD) HAMD-21 sum score at baseline was 20.8 ± 4.5 pts. (range: 10–31). Mean (±SD) sBDNF level at baseline was 22.5 ± 9.2 ng/ml (range: 7.0–51.4 ng/ml). Patients displayed “no sBDNF increase” from baseline to day 7 in 51%, and from baseline to day 14 in 34%. Patients were HAMD-21 non-improver on day 7 and day 14 in 55% and 43%, respectively. The combined marker non-increase of sBDNF plus non-improvement from baseline to day 7 or day 14 was present in 59% and 25%, respectively. 71% was non-responder at the endpoint, 81% was non-remitter. 3.2. Predictive value of early sBDNF non-increase for early HAMD-21 non-improvement The predictive values of a sBDNF non-increase from baseline to days 7 and 14 for HAMD improvement at days 7 and 14, respectively, are shown in Table 1. The sBDNF non-increase from BL to d7 predicted a HAMD-21 non-improvement from BL to d7 with moderate sensitivity and specificity; a sBDNF non-increase from BL to d14 predicted a HAMD-21 non-improvement from BL to d14 with low to moderate sensitivity and specificity. 3.3. Predictive value of early sBDNF non-increase for final non-response and non-remission The predictive values of the marker sBDNF non-increase from BL to days 7 and 14, respectively, for the treatment outcomes non-response and non-remission are shown in Table 2. A sBDNF non-increase from BL to days 7 and 14 predicted final non-response with 67% and 73% specificity, resp., and final non-remission with 63% and 71%, resp. specificity. Sensitivity for later non-response and non-remission was markedly lower. NPV for non-response (40 and 32%, resp.) and nonremission (25 and 20%, resp.) were far lower than the PPV (81 and 79%, resp.; 86 and 86%, resp). The area under the curve (ROC-AUC) for the prediction of a sBDNF change from baseline to days 7 and 14 on non-response at the endpoint was 60.3% (CI: 39.8–80.8%) and 59.7% (CI: 40.6–78.8%), respectively. The ROC-AUC for the prediction of a sBDNF change from baseline to days 7 and 14 on non-remission at the endpoint was 60.4% (CI: 35.6–84.2%) and 58.5% (CI: 37.1–79.9%), respectively. 3.4. Predictive value of early HAMD-21 non-improvement for final nonresponse and non-remission The predictive values of HAMD-21 non-improvement from BL to days 7 and 14, respectively, for the treatment outcomes non-response and non-remission are shown in Table 3. A HAMD-21 nonimprovement from baseline to days 7 and 14 was moderately to
418
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
Table 1 Predictive value of sBDNF non-increase from baseline to day 7 or 14 on HAMD-21 improvement on day 7 or 14, respectively. Predictor
Outcome (N) Non-Improvement d7
Improvement d7
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
57% (40.6–72.3)
56% (39.7–71.4)
62% (45.5–76.7)
50% (34.0–66.0)
1.63 (0.0–11.5)
44%
24% (11.8–40.3)
55% (38.3–70.9)
29% (15.6–45.7)
48% (31.8–64.6)
0.37 (0.0–9.7)
45%
sBDNF d7
Non-increase Increase
13 10
8 10
sBDNF d14a
Non-increase Increase
Non-Improvement d14 4 13
Improvement d14 10 12
CI: confidence interval; d: day; FDR: false discovery rate; non-improvement d7 (d14): HAMD-21 sum score decrease b20% between baseline and day 7 (14); NPV: negative predictive value. PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a N = 39 due to missing data of two patients at d14.
highly specific for later non-response (64 and 91%, resp.) and nonremission (57 and 86%, resp.). Sensitivity for later non-response and non-remission was lower. NPV for non-response and non-remission was far lower than the PPV (81 and 94%, resp.; 86 and 94%, resp). The ROC-AUC for the prediction of a HAMD-21 change from baseline to days 7 and 14 on non-response at the endpoint was 66.5% (CI: 44.5– 88.5%) and 85.4% (CI: 72.6–98.3%), respectively; the ROC-AUC for the prediction of a HAMD-21 change from baseline to days 7 and 14 on non-remission at the endpoint was 62.9% (CI: 35.2–90.6%) and 87.4% (CI: 00.0–100.0%), respectively.
3.5. Predictive value of an early sBDNF non-increase plus early HAMD-21 non-improvement for final non-response and non-remission The predictive values of the combined marker sBDNF non-increase plus HAMD-21 non-improvement from baseline to days 7 and 14, respectively, for the treatment outcomes non-response and nonremission are shown in Table 4. A sBDNF non-increase and HAMD-21 non-improvement from BL to days 7 and 14 were highly specific for later non-response (67 and 100%, resp.) and non-remission (60 and 100%, resp.). Sensitivity for later non-response and non-remission was lower. NPV for non-response and non-remission was far lower than the PPV (67 and 58%, resp.; 77 and 100%, resp.).
3.6. Determination of the most appropriate cut-off value for the prediction of treatment response at the endpoint by early sBDNF change ROC analysis revealed a sBDNF change of ± 4.5% between baseline and day 7 as optimal cut-off for the prediction of treatment response at EP; this value was nearly identical to the pre-defined cut-off of ± 3.7%.
4. Discussion Our pilot study provides evidence that the individual marker early sBDNF non-increase has a positive predictive value for final treatment failure very similar to the clinical marker early nonimprovement. This was not due to the fact that the non-increase of sBDNF is the major underlying biological factor of early-nonimprovement as documented by the additional analysis showing only a small overlap between early BDNF non-increase and early non-improvement. The non-increase of sBDNF has only a limited predictive value for early non-improvement, suggesting that sBDNF increase is only one of many other factors underlying the provocation of early improvement by antidepressant treatment. The biological reasons for the similar predictive values on treatment outcome are therefore unclear. It might be that the sBDNF change and the biological changes underlying early improvement are overlapping and lead to a common pathway underlying the course of symptom amelioration. In our study, the combined marker sBDNF non-increase plus early non-improvement has a higher predictive value compared to the individual markers sBDNF non-increase and early non-improvement: at week 2, the combined marker reaches a positive predictive value for treatment failure of 100%, thereby decreasing the rate of false negative predictions to 0%.Translated into clinical terms this means, the clinician could change the treatment strategy after two weeks of treatment, when he is observing a non-increase of sBDNF and a HAMD-21 non-improvement from treatment initiation to day 14, because no patient with these characteristics will experience a subsequent response or remission. In biological terms, the increase of the predictive value of the combined marker compared to the single markers again suggests different but overlapping biologicial entities of sBDNF increase on the one hand and the biology of early improvement on the other hand.
Table 2 Predictive value of sBDNF non-increase from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Predictor
sBDNF d7 sBDNF d14⁎
sBDNF d7 sBDNF d14a
Outcome (N) Non-Response
Response
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
Non-increase Increase Non-increase Increase
17 12 11 17
4 8 3 8
59% (42.6–74.1)
67% (50.6–80.9)
81% (65.7–91.5)
40% (25.0–56.5)
2.83 (0.1–13.5)
33%
39% (23.8–55.9)
73% (56.4–85.9)
79% (63.0–90.4)
32% (18.0–48.8)
1.73 (0.0–12.1)
25%
Non-increase Increase Non-increase Increase
Non-Remission 18 15 12 20
Remission 3 5 2 5
55% (38.7–70.6)
63% (46.5–77.5)
86% (71.6–94.8)
25% (12.8–41.0)
2.00 (0.0–12.1)
38%
38% (23.0–54.9)
71% (54.3–84.4)
86% (71.1–95.0)
20% (8.9–35.9)
1.50 (0.0–11.7)
28%
CI: Confidence interval; d: day; FDR: false discovery rate; non-response: HAMD-21 sum score decrease b50% between baseline and endpoint; non-remission: HAMD-21 sum score N 7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a N = 39 due to missing data of two patients at d14.
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
419
Table 3 Predictive value of HAMD non-improvement from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Predictor
HAMD d7 HAMD d14⁎
HAMD d7 HAMD d14a
Outcome (N) Non-Response
Response
Sensitivity (95% CL)
Specificity (95% CL)
PPV (95% CL)
NPV (95% CL)
Odds-Ratio (95% CL)
FDR
Non-improvement Improvment Non-improvement Improvment
19 10 16 13
4 8 1 10
66% (49.6–80.0)
67% (50.6–80.9)
83% (68.0–92.9)
44% (28.6–60.3)
3.80 (0.3–15.0)
33%
55% (38.5–70.7)
91% (77.6–97.7)
94% (81.7–99.0)
43% (27.5–59.6)
12.31 (4.0–26.8)
9%
Non-improvement Improvment Non-improvement Improvment
Non-Remission 20 13 16 17
Remission 3 5 1 6
61% (44.5–75.8)
63% (46.5–77.5)
87% (72.8–95.4)
28% (15.1–44.2)
1.85 (0.1–13.0)
38%
48% (32.0–64.3)
86% (71.4–94.9)
94% (81.7–99.0)
26% (13.4–42.3)
5.65 (0.8–17.8)
14%
CI: confidence interval; d: day; FDR: false discovery rate; Non-improvement d7 (d14): HAMD-21 sum score decrease b 20% between baseline and day 7 (14); Non-Response: HAMD21 sum score decrease b50% between baseline and endpoint; Non-Remission: HAMD-21 sum score N7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value. a N = 39 due to missing data of two patients at d14.
The results of this study suggest a valuable clinical utility of peripheral BDNF measurement as surrogate marker of antidepressant efficacy, as it has been proposed earlier (Brunoni et al., 2008; Sen et al., 2008; MacQueen, 2009). They provide further evidence for the importance of early changes of sBDNF in antidepressant treatment outcome and extend the existing results in both animals and humans: one study in humans investigated sBDNF levels at baseline and shortly after the initiation of antidepressant treatment (Gorgulu and Caliyurt, 2009). In this study, patients with MDD showed an increase of sBNDF within one day after sleep deprivation. Rats, which were treated with various antidepressant substances, showed a BDNF increase by posttranslational mechanisms within seven days in the hippocampus and the prefrontal-/frontal cortex (Musazzi et al., 2009). Taken together, the results support the hypothesis that sBDNF can be increased rapidly by the initiation of antidepressant treatment and – in turn – the absence of an early increase of sBDNF seems to indicate that the selected treatment will not be effective. It is not clear as to what extent peripheral BDNF reflects central BDNF (Sen et al., 2008; MacQueen, 2009). However, given that antidepressant treatment increases central BDNF expression (Nibuya et al., 1995; Malberg et al., 2000) and BDNF increase is associated with depressive symptom amelioration, the results in our study support
the idea of a close relation between peripheral and central BDNF. The explanation might be the existence of a bi-directional transport of BDNF crossing the blood-brain barrier, consisting of an active transport system for BDNF for the passage from blood to brain and the reabsorption from central spinal fluid to blood (Pan et al., 1998). One advantage of our pilot study is the repeated measurement of sBDNF levels starting shortly after the initiation of antidepressant treatment, because it enabled the investigation of the surrogate marker “sBDNF non-increase” in a time span of 7 and 14 days after treatment initiation. This could represent a progress in clinical decision making for the individual patient compared to the current practice of waiting for improvement during 4–6 weeks after treatment initiation. The results of our study with an association of sBDNF non-increase and treatment failure in numerous but not all individual patients of this study also show that BDNF is not able to fully encompass the complexity of MD. The close-meshed measurement of a biomarker of interest and the combination of biochemical and clinical markers might be useful for several other biological target systems presumably involved in the treatment response of major depression. The binarization of the sBDNF change from BL to days 7/14 as a predictor is a simplification which might have led to a loss of information, but mirrors what could be feasible and applicable in daily clinical practice. The herein defined
Table 4 Predictive value of sBDNF non-increase and HAMD-21 non-improvement from baseline to day 7 or 14 on HAMD-21 non-response and non-remission at the endpoint. Outcome (N)a
Predictor
Non-Response sBDNF + HAMD d7
sBDNF + HAMD d14
sBDNF + HAMD d7
sBDNF + HAMD d14
Non-increase + non-improvement Increase + improvement Non-increase + non-improvement Increase + improvement
Non-increase + non-improvement Increase + improvement Non-increase + non-improvement Increase + improvement
Response
Sensitivity (95% CI)
Specificity (95% CI)
PPV (95% CI)
NPV (95% CI)
Odds-Ratio (95% CI)
FDR
3
77% (54.3–92.0)
67% (43.9–85.3)
77% (54.3–92.0)
67% (37.7–79.6)
6.67 (0.5–25.9)
33%
44% (19.9–70.3)
100% (79.4–100)
100% (79.4–100)
58% (31.4–81.6)
0.00
65% (42.0–83.8)
60% (37.2–80.0)
85% (63.4–96.5)
33% (14.7–56.1)
2.75 (0.0–20.1)
33% (12.1–60.4)
100% (79.4–100)
100% (79.4–100)
33% (12.1–60.4)
0.00
6 4
0
5
7
Non-Remission 11
Remission 2
6
3
4
0
8
4
0%
40%
0%
CI: confidence interval; d: day; FDR: false discovery rate; non-improvement d7 (d14): HAMD-21 sum score decrease b20% between baseline and day 7 (14); non-response: HAMD21 sum score decrease b50% between baseline and endpoint; non-remission: HAMD-21 sum score N7 points at the endpoint; NPV: negative predictive value; PPV: positive predictive value; sBDNF: serum brain-derived neurotrophic factor. a Reduced N, because analysis included extreme groups, i.e. subjects with sBDNF non-increase plus HAMD non-improvement as well as sBDNF increase plus HAMD improvement; subjects with sBDNF non-increase plus HAMD improvement and sBDNF increase plus HAMD non-improvement were not incorporated in the analysis.
420
A. Tadić et al. / Progress in Neuro-Psychopharmacology & Biological Psychiatry 35 (2011) 415–420
cut-off has to be validated in future studies with healthy controls, larger samples of MD patients as well as in patients with different psychiatric disorders. Finally, the size of our sample is small and was associated with a false discovery rate between 0% and 45%; therefore, the results of our pilot study have to be interpreted cautiously and need validation in independent samples. 5. Conclusion In conclusion, the results of our study are promising in regard to the potential identification of a biomarker for treatment failure in patients with MDD which can be easily assessed in clinical routine. This is very important, because a biomarker for treatment nonresponse in MDD is warranted for both clinical application and industrial research. An independent replication of the results showing that the combined marker early-non-increase plus early-non-improvement predicts the final treatment failure to 100% could represent the basis for randomized controlled trials investigating its clinical value for the guidance of antidepressant treatment. Furthermore, the results of our study might provide a valuable model for the investigation of biomarkers in the treatment of MDD and other psychiatric disorders and might encourage research in this area in order to further shorten the duration of psychopharmacological treatment until the determination of insufficient effectiveness. Finally, these studies should lead to clear decision plans for guidance of psychopharmacological treatment of the individual patient. Financial disclosures None of the authors reported biomedical financial interests or potential conflicts of interest. Acknowledgements We are grateful to all participants for the support of study as well as to Dr. Tanja Zeller for the technical assistance. References American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders4th ed. . Washington DC: American Psychiatric Press; 2004. Brunoni AR, Lopes M, Fregni F. A systematic review and meta-analysis of clinical studies on major depression and BDNF levels: implications for the role of neuroplasticity in depression. Int J Neuropsychopharmacol 2008;11:1169–80. Duman RS, Monteggia LM. A neurotrophic model for stress-related mood disorders. Biol Psychiatry 2006;59:1116–27. Fournier JC, DeRubeis RJ, Hollon SD, Dimidjian S, Amsterdam JD, Shelton RC, et al. Antidepressant drug effects and depression severity: a patient-level meta-analysis. JAMA 2010;303:47–53. Frank E, Prien RF, Jarrett RB, Keller MB, Kupfer DJ, Lavori PW, et al. Conceptualization and rationale consensus definitions of terms in major depressive disorder. Remission, recovery and recurrence. Arch Gen Psychiatry 1991;48:851–5.
Gorgulu Y, Caliyurt O. Rapid antidepressant effects of sleep deprivation therapy correlates with serum BDNF changes in major depression. Brain Res Bull 2009;80:158–62. Henkel V, Seemüller F, Obermeier M, Adli M, Bauer M, Mundt C, et al. Does early improvement triggered by antidepressants predict response/remission? Analysis of data from a naturalistic study on a large sample of inpatients with major depression. J Affect Dis 2009;115:439–49. Hennings JM, Owashi T, Binder EB, Horstmann S, Menke A, Kloiber S, et al. Clinical characteristics and treatment outcome in a representative sample of depressed inpatients. Findings from the Munich Antidepressant Response Signature (MARS) project. Psychiatric Res 2009;43:215–29. Holsboer F. How can we realize the promise of personalized antidepressant medicines? Nat Rev Neurosci 2008;9:638–46. Katz MM, Tekell JL, Bowden CL, Brannan S, Houston JP, Berman N, et al. Onset and early behavioral effects of pharmacologically different antidepressants and placebo in depression. Neuropsychopharmacology 2004;29:566–79. Kirsch I, Deacon BJ, Huedo-Medina TB, Scoboria A, Moore TJ, Johnson BT. Initial severity and antidepressant benefits: a meta-analysis of data submitted to the Food and Drug Administration. Public Libr Sci Med 2008;5:e45. Krishnan V, Nestler EJ. The molecular neurobiology of depression. Nature 2008;455: 894–902. Leuchter AF, Cook IA, Hunter AM, Korb AS. A new paradigm for the prediction of antidepressant treatment response. Dialogues Clin Neurosci 2009;11:435–46. MacQueen GM. Magnetic resonance imaging and prediction of outcome in patients with major depressive disorder. J Psychiatry Neurosci 2009;34:343–9. Malberg JE, Eisch AJ, Nestler EJ, Duman RS. Chronic antidepressant treatment increases neurogenesis in adult rat hippocampus. J Neurosci 2000;20:9104–10. Musazzi L, Cattaneo A, Tardito D, Barbon A, Gennarelli M, Barlati S, et al. Early raise of BDNF in hippocampus suggests induction of posttranscriptional mechanisms by antidepressants. Biomed Cent Neurosci 2009;10:48. Nibuya M, Morinobu S, Duman RS. Regulation of BDNF and trkB mRNA in rat brain by chronic electroconvulsive seizure and antidepressant drug treatments. J Neurosci 1995;15:7539–47. Nierenberg AA, Farabaugh AH, Alpert JE, Gordon J, Worthington JJ, Rosenbaum JF, et al. Timing of onset of antidepressant response with fluoxetine treatment. Am J Psychiatry 2000;157:1423–8. Pan W, Banks WA, Fasold MB, Bluth J, Kastin AJ. Transport of brain-derived neurotrophic factor across the blood-brain barrier. Neuropharmacology 1998;37:1553–61. Sen S, Duman R, Sanacora G. Serum brain-derived neurotrophic factor, depression, and antidepressant medications: meta-analyses and implications. Biol Psychiatry 2008;64:527–32. Sheehan DV, Lecrubier Y. M.I.N.I. International Neuropsychiatric Interview. USA Tampa; 1998. Stassen HH, Angst J, Hell D, Scharfetter C, Szegedi A. Is there a common resilience mechanism underlying antidepressant drug response? Evidence from 2848 patients. J Clin Psychiatry 2007;68:1195–205. Szegedi A, Müller MJ, Anghelescu I, Klawe C, Kohnen R, Benkert O. Early improvement under mirtazapine and paroxetine predicts later stable response and remission with high sensitivity in patients with major depression. J Clin Psychiatry 2003;64:413–20. Szegedi A, Jansen WT, van Willigenburg AP, van der Meulen E, Stassen HH, Thase ME. Early improvement as a predictor of treatment outcome in patients with major depressive disorder: why the first 2 weeks really matter-evidence from 6,562 patients. J Clin Psychiatry 2009;70:344–53. Tadić A, Helmreich I, Mergl R, Hautzinger M, Henkel V, Hegerl U. Early improvement is a predictor of treatment outcome in patients with mild major, minor or subsyndromal depression. J Affect Dis 2010a;120:86–93. Tadić A, Gorbulev S, Dahmen N, Hiemke C, Braus DF, Röschke J, et al. EMC Study Group. Rationale and design of the randomised clinical trial comparing early medication change (EMC) strategy with treatment as usual (TAU) in patients with major depressive disorder—the EMC trial. Trials 2010b;11:21. Turner EH, Matthews AM, Linardatos E, Tell RA, Rosenthal R. Selective publication of antidepressant trials and its influence on apparent efficacy. N Engl J Med 2008;358: 252–60. Warden D, Rush AJ, Trivedi MH, Fava M, Wisniewski SR. The STAR*D Project results: a comprehensive review of findings. Curr Psychiatry Rep 2007;9:449–59.