Percentage reduction of depression severity versus absolute severity after initial weeks of treatment to predict final response or remission

Authors

  • Ching-Hua Lin MD, PhD,

    1. Kaohsiung Municipal Kai-Syuan Psychiatric Hospital, Kaohsiung, Taiwan
    2. Graduate Institute of Medicine, College of Medicine, Kaohsiung Medical University, Kaohsiung, Taiwan
    Search for more papers by this author
  • Cheng-Chung Chen MD, PhD,

    1. Kaohsiung Municipal Kai-Syuan Psychiatric Hospital, Kaohsiung, Taiwan
    Search for more papers by this author
  • Fu-Chiang Wang MD,

    1. Kaohsiung Municipal Kai-Syuan Psychiatric Hospital, Kaohsiung, Taiwan
    Search for more papers by this author
  • Hsien-Yuan Lane MD, PhD

    Corresponding author
    1. Institute of Clinical Medical Science, China Medical University, Taichung, Taiwan
    2. Departments of Psychiatry, China Medical University Hospital, Taichung, Taiwan
    • Kaohsiung Municipal Kai-Syuan Psychiatric Hospital, Kaohsiung, Taiwan
    Search for more papers by this author

Correspondence: Hsien-Yuan Lane, MD, PhD, Department of Psychiatry, China Medical University Hospital (Taiwan), no. 2, Yuh-Der Road, Taichung City 404, Taiwan. Email: hylane@pchome.com.tw; hylane@gmail.com

Abstract

Aim

Percentage reduction of depression severity has been used to predict both response and remission of major depression. We aimed to compare the accuracy to predict response or remission by percentage reduction of depression score or depression score after initial weeks of treatment.

Methods

The subjects were 126 depressed inpatients who received 20 mg/day fluoxetine for 6 weeks. Symptom severity was assessed using the 17-item Hamilton Depression Rating Scale (HAMD-17). Response was defined as a reduction of 50% or more of the HAMD-17. Remission was defined as a score of ≤7 of the HAMD-17. At weeks 1, 2, 3 and 4, the percentages of HAMD-17 score reduction, the percentages of mood cluster score reduction, HAMD-17 scores, and mood cluster scores were regarded as potential predictors. The receiver operating characteristic curve was applied to determine the cut-off point of predictors at weeks 1, 2, 3, and 4.

Results

One-hundred and seven patients completed the 6-week trial. At weeks 1, 2, 3, and 4, percentages of HAMD-17 score reduction or HAMD-17 scores were the best predictors of responder or remitters, respectively. Using the percentage of HAMD-17 score reduction at each assessment as a predictor of response generated a larger area under the curve than other predictors. Conversely, applying the absolute HAMD-17 score at each assessment as a predictor of remission had the largest area under the curve.

Conclusion

Applying percentage of reduction in depression severity during the early weeks of treatment can predict response, and it is reasonable to apply depression severity to predict remission.

CLINICAL PREDICTION MODELS assist clinicians in making early decisions in treating major depressive disorder (MDD). They are comprised of clinical variables that show strong predictive values for outcome.[1] Early prediction for poor outcome to antidepressants could prevent unnecessary persistence with ineffectual agents, diminish risk of adverse events, decrease duration of depressive episodes, and reduce the cost and burden of illness.

In clinical trials for antidepressants, response is a widely used outcome variable,[2] and is usually defined as a score reduction of 50% or more in standardized depression scales, such as the 17-item Hamilton Depression Rating Scale (HAMD-17). However, a depressed patient who responds with a 50% reduction of the HAMD-17 score may still manifest significant symptoms, particularly if he/she was severely depressed prior to treatment (for example, a baseline HAMD-17 score of 34). Remission is defined according to post-treatment scores of a depression rating scale, and is commonly defined as a low absolute score of ≤7 on the HAMD-17.[3] However, response is not synonymous with remission.[4]

Several studies[5-9] used a certain percentage of improvement (e.g. 20, 25, or 30%) from baseline by a depression scale during the early weeks to predict response or remission for depressed patients undergoing acute treatment, and suggested that percentage improvement in the early weeks can predict subsequent response or remission.

This study aimed to find accurate methods to make an early prediction of final response or remission of MDD patients by comparing the percentage reduction of depression severity and absolute depression severity after the initial weeks of acute treatment.

Methods

Subjects

The study was approved by the Kai-Syuan Psychiatric Hospital's institutional review board and conducted in accordance with Good Clinical Practice procedures and the current revision of the Declaration of Helsinki (Project number: KSPH-2007-16). Written, informed consent was obtained from the participants. This study has also been registered on Clinical.trials.gov (Identifier number: NCT01075529).

This study was a post-hoc analysis. Details of the patient sample have been presented elsewhere.[10, 11] In brief, subjects were recruited from Kai-Syuan Psychiatric Hospital. Participants were considered eligible at admission if they were: newly hospitalized patients undergoing acute treatment, between 18 and 70 years old, physically healthy and had all laboratory parameters within normal limits, taking antidepressants other than fluoxetine or no antidepressant use within 1 week before admission, and diagnosed for an MDD using the Structured Clinical Interview for DSM-IV (SCID).[12] The exclusion criteria included a 17-item Hamilton Depression Rating Scale (HAMD-17) < 18 and a Clinical Global Impression of Severity (CGI-S)[13] <4 at baseline, psychotic depression, bipolar I or II disorder, schizophrenia or any other psychotic disorder, a DSM-IV diagnosis of substance abuse or dependence (including alcohol) within the past 6 months, mental disorders due to organic factors, severe cognitive impairment, initiating or ending formal psychotherapy within 6 weeks prior to enrollment, treatment-resistant depression (defined as a lack of response to two or more adequate courses of antidepressant treatment), a history of poor response to fluoxetine (20 mg/day for ≥4 weeks), a history of electroconvulsive therapy, pregnancy or lactation, and exposure to fluoxetine within 1 month before admission.

Procedures

After a washout period of at least 72 h since admission, patients received open-label fluoxetine treatment at fixed doses of 20 mg daily for 6 weeks.[14] During the course of treatment, psychiatrists had the option of adding certain anxiolytic and/or sedative-hypnotic medications for brief periods, as determined by clinical necessity. No other psychotropic agents were allowed. Drug adherence was monitored and ensured by psychiatric nurses.

Symptom severity was assessed at baseline, and again at weeks 1, 2, 3, 4 and 6 by trained, qualified psychiatrists using HAMD-17. HAMD-17 scores range from 0 to 52, with higher scores indicating more severe depression. The intraclass correlation coefficient of reliability was 0.95 between the raters. To maintain high interrater reliability and prevent rater drift, raters met at least once a month for training and reliability re-testing. To ensure objectivity, the research psychiatrists conducting the clinical ratings did not know the detailed study design.

Response was defined as a reduction of 50% or more of the HAMD-17 score after 6 weeks of treatment. Remission was defined as a score of ≤7 on the HAMD-17 after 6 weeks of treatment.[3] Thus, response status or remission status was a dichotomous variable, operationally defined as ‘response/non-response’ or ‘remission/non-remission’, respectively.

Statistical analyses

To exclude anxiety symptoms in HAMD-17, a mood cluster with seven items was regarded as core depressive symptoms.[15] These seven items are: (1) depressed mood; (2) feelings of guilt; (3) suicide; (7) work and activities; (8) psychomotor retardation; (13) somatic symptoms, general; and (14) genital symptoms, sexual interest.

At weeks 1, 2, 3 and 4, the percentages of HAMD-17 score reduction, the percentages of mood cluster score reduction, HAMD-17 scores, and mood cluster scores were regarded as potential predictors. If potential predictors were found to be significant by means of statistical analysis, they were entered into a receiver operating characteristic (ROC) curve to determine both the best model and the best predictor. ROC curve was used to determine the best cut-off point of the predictor between responders and non-responders or between remitters and non-remitters, and to maximize both the sensitivity and specificity of the predictor variable so that false positive and false negative rates could be minimized. The ROC curve is created by plotting 1-specificity against sensitivity to distinguish response from non-response or remission from non-remission. The area under the ROC curve (AUC) is used to quantify the ability or accuracy of the test in identifying responders from non-responders or remitters from non-remitters. In practice, an AUC generally falls somewhere between 0.50 and 1, with an AUC of 0.8 considered as having a good discriminative capacity.[16, 17] All tests were two-tailed, and significance was defined as an alpha of less than 0.05. All data were processed by SPSS version 17.0 for Windows (SPSS, Chicago, IL, USA) and MedCalc (MedCalc Software, Mariakerke, Belgium). MedCalc is a program that implements several statistical procedures, including ROC analysis.

Results

Patients

A total of 126 acutely ill inpatients with MDD were enrolled. One-hundred and seven (84.9%) of the 126 patients completed the 6-week trial of fluoxetine (completers). Twenty-one (19.6%) of the completers were male and 86 (80.4%) were female. The mean age was 46.2 ± 10.8 years. Eighty-two of the completers did not take any antidepressants within 1 week before admission (data not shown). The remaining 19 did not complete the trial due to: lack of efficacy (n = 3), premature discharge (n = 14) or withdrawal of consent (n = 2). No patient withdrew from the study due to adverse events.

Response versus remission

Of the 107 completers, 58.9% (n = 63) or 27.1% (n = 29) of the subjects were classified as responders or remitters after the 6-week treatment. There were no significant differences between responders and non-responders in terms of sex (P = 0.07) and age (46.9 ± 9.7 vs 45.1 ± 12.1, P = 0.40). Remitters and non-remitters were comparable for sex (P = 0.87) and age (47.4 ± 11.2 vs 45.8 ± 10.7, P = 0.49). At weeks 1, 2, 3 and 4, percentages of HAMD-17 score reduction, percentages of mood cluster score reduction, HAMD-17 scores, and mood cluster scores were valid predictors of responders or remitters (see Tables 1 and 2).

Table 1. Early predictors at weeks 1, 2, 3, and 4 for response (≥50% reduction in HAMD-17 at week 6)
Early predictorResponders (n = 63)Non-responders (n = 44)Pa
MeanSDMeanSD
  1. Bold, statistically significant.
  2. aIndependent t-test.
  3. bMood cluster = items 1, 2, 3, 7, 8, 13, and 14 of HAMD-17.
  4. HAMD-17 = 17-item Hamilton Depression Rating Scale.
Percentage of HAMD-17 score reduction     
Week 137.720.122.718.6<0.001
Week 253.320.429.617.9<0.001
Week 360.219.030.419.8<0.001
Week 467.315.929.617.2<0.001
Percentage of mood clusterb score reduction     
Week 135.022.122.520.50.004
Week 252.425.029.019.3<0.001
Week 359.823.529.324.2<0.001
Week 467.520.728.121.5<0.001
HAMD-17 score     
Week 031.66.832.46.40.53
Week 119.67.825.07.90.001
Week 214.97.722.66.6<0.001
Week 312.87.422.47.1<0.001
Week 410.45.722.76.1<0.001
Mood cluster score     
Week 015.03.815.63.00.45
Week 110.04.612.14.00.017
Week 27.54.711.03.4<0.001
Week 36.44.411.04.1<0.001
Week 45.13.511.33.9<0.001
Table 2. Early predictors at weeks 1, 2, 3, and 4 for remission (HAMD-17 score ≦7 at week 6)
Early predictorRemitters (n = 29)Non-remitters (n = 78)P
MeanSDMeanSD
  1. Bold, statistically significant.
  2. HAMD-17 = 17-item Hamilton Depression Rating Scale.
Percentage of HAMD-17 score reduction     
Week 143.817.427.020.2<0.001
Week 260.119.437.420.6<0.001
Week 367.218.940.822.0<0.001
Week 477.513.342.321.0<0.001
Percentage of mood cluster score reduction     
Week 141.919.725.421.5<0.001
Week 259.723.936.523.2<0.001
Week 368.622.039.326.0<0.001
Week 479.618.740.824.2<0.001
HAMD-17 score     
Week 029.27.633.06.00.008
Week 116.26.723.97.8<0.001
Week 211.67.020.57.3<0.001
Week 39.66.619.47.7<0.001
Week 46.34.018.87.0<0.001
Mood cluster score     
Week 013.54.015.93.00.001
Week 18.14.111.94.2<0.001
Week 25.63.910.24.2<0.001
Week 34.43.69.74.5<0.001
Week 42.82.69.44.1<0.001

Early prediction

The cut-off point, sensitivity, specificity, and AUC of each valid predictor using ROC analysis are shown in Tables 3 and 4. At weeks 1, 2, 3 and 4, HAMD-17 score reductions of 27%, 37%, 40%, and 45% seemed to be the optimal cut-off points for predicting eventual response. This analysis provided an AUC of 72%, 81%, 86%, and 96% (Table 3). In contrast to response, HAMD-17 scores of 22, 14, 12, and 8 also seemed to be the optimal cut-off points for predicting eventual remission. This analysis provided an AUC of 77%, 82%, 84%, and 95% (Table 4).

Table 3. Prediction of response by percentages of HAMD-17 score reduction, percentages of mood cluster score reduction, HAMD-17 score, and mood cluster score at weeks 1, 2, 3, and 4: ROCa analysis
Early predictorsCut-off pointSensitivitySpecificityAUC
  1. aFor determining the cut-off point of valid predictors by plotting the proportion of true-positive results (sensitivity) versus the proportion of false-positive results (1 – specificity).
  2. AUC, area under the ROC curve (100%); HAMD-17, 17-item Hamilton Depression Rating Scale; ROC, receiver operating characteristic.
Percentage of HAMD-17 score reduction    
Week 127%71%71%72%
Week 237%84%75%81%
Week 340%83%75%86%
Week 445%92%89%96%
Percentage of mood cluster15 score reduction    
Week 116%83%50%69%
Week 247%54%89%77%
Week 346%73%82%81%
Week 445%83%84%91%
HAMD-17 score    
Week 11854%80%69%
Week 21660%86%78%
Week 31362%91%83%
Week 41681%84%92%
Mood cluster score    
Week 11164%64%64%
Week 2648%93%74%
Week 3767%80%79%
Week 4883%77%87%
Table 4. Prediction of remission by percentages of HAMD-17 score reduction, percentages of mood cluster score reduction, HAMD-17 score, and mood cluster score at weeks 1, 2, 3, and 4: ROC analysis
Early predictorsCut-off pointSensitivitySpecificityAUC
  1. AUC, area under the ROC curve (100%); HAMD-17, 17-item Hamilton Depression Rating Scale; ROC, receiver operating characteristic.
Percentage of HAMD-17 score reduction    
Week 129%83%63%75%
Week 243%90%62%79%
Week 364%72%86%82%
Week 472%75%89%93%
Percentage of mood cluster score reduction    
Week 127%79%67%74%
Week 247%69%76%76%
Week 361%72%83%81%
Week 466%83%89%90%
HAMD-17 score    
Week 12283%64%77%
Week 21479%78%82%
Week 31283%82%84%
Week 4883%94%95%
Mood cluster score    
Week 11072%68%74%
Week 2666%82%79%
Week 3569%85%82%
Week 4479%90%92%

Discussion

The main finding of this study is that using percentage of HAMD-17 score reduction at weeks 1, 2, 3, or 4 as a predictor for response generated a larger AUC than using percentage of mood cluster score reduction, absolute HAMD-17 score, or absolute mood cluster score at weeks 1, 2, 3, or 4, respectively (see Table 3). Meanwhile, applying an absolute HAMD-17 score at weeks 1, 2, 3, or 4 as a predictor for remission generated a larger AUC than applying percentage of HAMD-17 score reduction, percentage of mood cluster score reduction, or mood cluster score at weeks 1, 2, 3, or 4, respectively (see Table 4).

The second finding is that the best cut-off point of predictors (i.e. percentage of HAMD-17 score reduction for response or HAMD-17 scores for remission), determined by ROC curves at weeks 2, 3 and 4 (Tables 3 and 4), had adequate discriminatory ability (i.e. an AUC ≥ 0.8)[16] in identifying final responders or remitters at the end of the 6-week treatment.

There are several advantages for busy physicians to predict final remission using symptom severity rather than percentage symptom reduction during the early weeks after treatment. First, it is user-friendly to spare the relatively complex calculation to obtain the percentage symptom reduction. Second, while baseline symptom severity is not easy or not reliable to be obtained from a severely depressed patient who may be incapable of clearly revealing his/her mental status at the initial visit, a clinician is still able to use depression severity at week 2, 3, or 4 to predict ultimate remission. Finally, as aforementioned, using depression severity to predict remission is more accurate.

According to two studies[6, 7] and our previous report,[11] stable response was defined as a reduction of 50% or more of the HAMD-17 score at weeks 4 and 6 of treatment. Stable remission was defined as an HAMD-17 score ≤ 7 at weeks 4 and 6. We also found that both the percentages of HAMD-17 score reduction at weeks 1, 2, 3 and 4, and HAMD-17 scores at weeks 1, 2, 3 and 4 were valid predictors of stable responders or stable remitters. At weeks 1, 2, 3 and 4, we also found that the percentages of HAMD-17 score reduction, the percentages of mood cluster score reduction, HAMD-17 scores, and mood cluster scores were valid predictors of stable responders or stable remitters. Percentages of HAMD-17 score reduction at weeks 1, 2, 3 and 4 had the larger AUC (78%, 86%, 90%, and 99%) than HAMD-17 scores at weeks 1, 2, 3 and 4 (75%, 83%, 87%, and 95%) or other predictors (data not shown) to predict final stable response. Meanwhile, HAMD-17 scores at weeks 1, 2, 3 and 4 provided the larger AUC (78%, 82%, 86%, and 98%) than percentages of HAMD-17 score reduction at weeks 1, 2, 3 and 4 (75%, 79%, 84%, and 96%) or other predictors to predict final stable remission (data not shown).

Tedlow et al.[18] have suggested that various treatment outcomes need different predictors. Response or stable response is defined by percentage of reduction in depression severity. Remission or stable remission is defined by residual depression symptoms. Applying percentage of reduction in depression severity during the early weeks to predict final response or applying depression severity to predict final remission is reasonable.

The remission rate (27.1%) and response rate (58.9%) in this study are comparable to the rates in a pooled analysis, which had a 29% remission rate and a 59% response rate.[19] Patients in the current study received the same fixed dose of 20 mg/day of fluoxetine. Earlier fixed-dose studies[14, 20] have demonstrated that 20 mg of fluoxetine daily is the optimal dose for most patients, and is associated with fewer and less severe side-effects than higher doses. A meta-analysis study[21] also suggests that fluoxetine therapy at 20 mg daily is an adequate therapy with good tolerance. Another system review[22] found no evidence of increased efficacy with SSRI's escalation from medium to high dosages.

Several strengths of this study should be addressed. First, the subjects were inpatients requiring acute treatment. Hospitalized patients constituted only a small proportion of the patients in previous studies.[23] While hospitalized, the subjects were carefully monitored for symptom assessment, the development of side-effects, and medical adherence. They also had similar environments. Second, inpatients experience greater severity of depression than outpatients. It has been reported that patients with more severe depression get fewer placebo effects.[24] Third, HAMD-17 was originally developed for inpatients,[25] and might therefore be sufficiently, or even more sensitive in detecting changes in depressive symptoms of greater severity.[26, 27] Finally, unlike other studies which used last-observation-carried-forward (LOCF) analysis to account for missing data, we analyzed only the trial completers. LOCF analysis, assuming that a subject's severity rating at the time of dropout would be the same as his/her rating at the end of the trial, could add a negative bias to the results.[28]

The limitations of this study included a short-term, open-labeled treatment design, short-term washout period, the use of a single antidepressant agent (fluoxetine), and a single-center study. The ideal duration for an antidepressant trial remains controversial.[29] Stassen et al.[30] have indicated that extending clinical trials of antidepressants beyond 6 weeks does not yield more information. However, the traditional period of 6 weeks may not be the most appropriate, given that a subset of patients go on to achieve remission after 6 weeks.[27] One difficulty of the current study is that most patients refused to stay in hospital for longer than 6 weeks. Moreover, it would be unethical to continue an ineffective treatment for more than 6 weeks. Regarding the study design, this open-labeled study, albeit not as precise as a double-blind study, may have been adequate for early identification of responder or remitters. Additionally, an open trial more closely resembles the clinical practice setting,[31] in which patients and clinicians both know the medication and both expect the outcomes. Consequently, the results may be more generalizable to clinical settings. Furthermore, the subjects were inpatients who had been hospitalized due to severe symptoms, severe functional impairment, or suicidal tendencies. Thus, inclusion of a placebo group would have given rise to ethical concerns. Considering the short washout period, this study excluded the patients who had taken fluoxetine, which has a long half-life, within 1 month before admission. A long washout period would be another ethical concern for newly hospitalized patients.

It is also unclear whether the conclusions could be generalized to other antidepressants or to outpatients. A major difference between inpatients and outpatients is that adherence to protocol treatment can be ascertained during hospitalization. If a patient would maintain a good medical adherence after discharge, we assume that the findings of our study could be generalizable to him/her. Further studies, preferably involving other antidepressants or cognitive behavioral therapy, larger inpatient or outpatient groups from multi-centers, and duration of longer than 6 weeks, are needed.

Acknowledgments

This study was funded by the Kai-Syuan Psychiatric Hospital in 2007–2009, the Food and Drug Administration, Department of Health, Taiwan (DOH100-FDA-43002-L), the National Science Council, Taiwan (NSC-101–2325-B-039-009, NSC-101–2314-B-039-030-MY3, and NSC 99–2314-B-280 −001-MY3), the National Health Research Institutes, Taiwan (NHRI-EX-101-9904NI), Taiwan Department of Health Clinical Trial and Research Center of Excellence (DOH102-TD-B-111-004), and China Medical University Hospital, Taiwan (CMU 101-AWARD-13, DMR-100-118, and DMR-100-185). All authors declare that they have no conflicts of interest. This trial was registered with Clinical.trials.gov (no. NCT01075529).

Ancillary