Initial studies-such as the stepped collaborative care intervention, Texas Medication Algorithm Project (TMAP), and German Algorithm Project (GAP) phase 2-predominantly investigated whether following an expert opinion–based clinical algorithm (irrespective of the content of the algorithm) led to a better outcome than treatment as usual did
In my last 2 columns, I discussed the problems associated with clinical practice guidelines and described how some physicians use them in practice. In this column, I will review studies on the validation of clinical algorithms, which are 1 of the 2 main types of practice guidelines. Algorithms are detailed, step-by-step flow charts that outline the recommended treatment for patients with a specific disorder.1
Initial studies-such as the stepped collaborative care intervention, Texas Medication Algorithm Project (TMAP), and German Algorithm Project (GAP) phase 2-predominantly investigated whether following an expert opinion-based clinical algorithm (irrespective of the content of the algorithm) led to a better outcome than treatment as usual did.2-4 Subsequent studies, including the Sequenced Treatment Alter na tives to Relieve Depression (STAR*D) and GAP phase 3, focused on comparing the effectiveness of different treatments that were provided at each step in a sequenced algorithm.5,6 The former studies tell us something about the process of care, or how treatment should be provided. The latter tell us about the content of care, or which treatment should be provided at each stage in the algorithm.
Clinical algorithm studies
Several studies suggest that patients with a major depressive or bipolar disorder who are treated according to an accepted clinical algorithm (ALGO) have more successful outcomes than those treated as usual (TAU).6 Some of these studies, and their results, are discussed below.
Stepped collaborative care study
Katon and colleagues3,7 studied the effect of a stepped collaborative care intervention for primary care patients who had persistent or treatment-resistant depression that did not respond to several weeks of conventional antidepressant treatment. Patients were divided into 2 groups: an ALGO group, treated using a "stepped-care" algorithm that included education and structured clinical interventions, and a TAU group, who were simply told to speak with their primary care physician about treatment for depression. Depression was measured using the 20-question depression subscale of the Hopkins Symptom Checklist-90. Patients in the ALGO group who had moderately severe depression-but not those who had severe depression-were significantly improved compared with patients in the TAU group (P = .004) after several months of treatment.
In the GAP, Adli and colleagues8 developed a standardized stepwise drug treatment regimen (SSTR) for treatment-resistant depression. In a randomized controlled study, they used the Bech-Rafaelsen Melancholia Scale (BRMS) to compare the outcome of patients treated by SSTR (n = 74) and TAU (n = 74).4 SSTR-treated patients had a better outcome than TAU patients (BRMS scores: SSTR = 5.4, TAU = 9.5; P < .01). The SSTR group also had a significantly higher dropout rate (45%) than the TAU group (16%). The authors commented that 33% of the dropouts were due to physician non compliance with algorithm rules.
Researchers in the TMAP developed detailed algorithms for the treatment of major depressive disorder, schizophrenia, and bipolar disorder in the Texas public health system. In a 12-month study of the effectiveness of the algorithm in the treatment of major depressive disorder, Trivedi and associates2 assigned patients with depression to an ALGO group (n = 175) or a TAU group (n = 175). After 3 months of treatment, ALGO patients had significantly fewer depressive symptoms (P = .002), as measured on the Inventory of Depressive Symptomatology-Clinician Rating Scale (IDS-C30), than the TAU group. During the remaining 9 months of treatment, there was no difference in the rate of improvement of the 2 groups (P = .74), but TAU patients never caught up with the initial lead established by the ALGO group.
ALGO treatment, however, does not always produce a significantly better response than TAU. Suppes and col leagues9 studied the clinical response of a group of patients with bipolar disorder who were treated according to the TMAP bipolar algorithm. In the 12-month study, the ALGO group showed significant improvements (P = .03) on the Brief Psychiatric Rating Scale (BPRS-24) compared with the TAU group in the first 3 months. The differences between the 2 groups narrowed considerably from months 3 to 12 as the TAU group's rate of improvement on the scale increased.
Manic and hypomanic symptoms, measured by the Clinician-Adminis tered Rating Scale for Mania, decreased significantly more in the ALGO group (P = .007) than in the TAU group during the first 3 months of treatment. The ALGO patients continued to do better than the TAU patients from months 3 to 12, but there was no significant difference in the rate of improvement between the 2 groups. ALGO patients demonstrated a significantly greater decrease in psychotic symptoms over TAU patients in the first 3 months of treatment, but the TAU group caught up by month 12. There were no significant differences in the symptoms of depression, measured on the IDS-C30, between the ALGO and TAU groups during the 12 months of the study.
In a second TMAP bipolar study, investigators examined the relationship between the clinical response of patients to treatment and the adherence of their providers to the TMAP bipolar algorithm.10 The authors used a complex formula to calculate the amount of improvement experienced by ALGO-treated and TAU patients. ALG patients who had depression demonstrated statistically significant improvements (P = .005) in symptoms on the IDS-C30 compared with TAU patients during the last 9 months of the 12-month study. The ALGO group also demonstrated greater improvement than the TAU group on the BPRS-24 during that time. ALGO treatment, however, did not produce significant improvements in manic or hypomanic symptoms over TAU. Curiously, providers who had more experience with the TMAP algorithm adhered to it less.
The STAR*D trial was designed to objectively determine an effective 4-level sequence of treatments for outpatients with nonpsychotic, treatment-resistant depression.5 In an initial STAR*D study, 2876 psychiatric and primary care outpatients with depression were treated at level 1 with citalopram (Celexa).11 Depression was mea sured using the 17-item Hamilton De pression Rating Scale (HAM-D) and the 16-item Quick Inventory of Depressive Symptomatology, Self-Report. By the end of the study, 790 (27.5%) patients had achieved remission (de fined as a score of 7 or less on the HAM-D). A large proportion of these patients required 8 or more weeks of treatment with an average dosage of 41.8 mg/d of citalopram.
In a subsequent STAR*D trial, 727 patients who did not respond to level 1 treatment were randomly assigned to level 2 treatment with sustained-release bupropion (Wellbutrin), sertraline (Zoloft), or extended-release venlafaxine (Effexor).12 About one quarter of these patients achieved remission. The difference in response to the 3 medications was not statistically significant. A second group of 851 patients who did not achieve remission with citalopram in the level 1 trial received level 2 augmentations with sustained-release bupropion or buspirone (BuSpar). The proportion of patients who achieved remission was similar for the 2 medications (29.7% for sustained-release bupropion and 30.1% for buspirone).13
Fava and associates14 treated 235 patients who had not responded to level 1 or 2 treatments in either STAR*D trial with mirtazapine (Remeron) or nor triptyline (Aventyl, Pamelor) for 14 weeks (level 3). Remission rates were 12.3% for mirtazapine and 19.8% for nortriptyline. The differences in out come between the 2 treatments were not statistically significant.
What conclusions can we draw from these comprehensive, well-designed, and well-executed studies? Like most complex investigations, they resolve some important questions and raise others. The collaborative care, TMAP, and GAP studies repeatedly indicated that ALGO treatment produces a statistically better clinical response than TAU. Unfortunately, it is not exactly clear what this means because we do not know any of the details of how the TAU patients were treated. For example, it would be useful to know whether physicians who treated patients without algorithms made specific types of mistakes that physicians who did use them avoided. Did they routinely treat patients with too little medication for too short a length of time? If so, those mistakes might easily be corrected. In any case, it would be useful to study the TAU group to determine what errors in treatment decreased the potential effectiveness of their care.
Investigators have suggested 2 possible reasons for the improved outcomes seen in patients treated with algorithms: it could be the content of the ALGO treatments or the diligent, organized man ner in which they are provided to patients.6 The former seems less likely since the sequence and-to some extent-the content of each step in the TMAP and GAP algorithms were different, yet each algorithm demonstrated superior effectiveness over TAU.
The second and more likely reason for the superiority of ALGO treatment is the consistent and organized manner in which care was provided. Trivedi and colleagues11 referred to this diligent care as "measurement-based care" and defined it as "the routine mea sure ment of symptoms and side effects at each treatment visit" followed by an appropriate adjustment of medication. It is possible that the simple act of following the algorithm, like a to-do list, continually reminds the physician of the necessary components of care. It is also possible that there were more resources devoted to ALGO patients and that the use of these resources was responsible for the patients' higher rate of remission.
The STAR*D studies tell us something interesting about the choice of specific treatments at each level in the care of patients with treatment-resistant depression-the treatments are apparently equivalent. It made little difference in level 2 treatment whether a patient was given sustained-release bupropion, sertraline, or extended-release venlafaxine or whether the initial citalopram treatment was augmented with sustained-release bupropion or buspirone. The same was true in level 3 treatment, where mirtazapine and nortriptyline produced similar responses. The results suggest that clinicians could have given patients sus tained-release bupropion or sertraline as the initial medicine, rather than citalopram, and obtained the same results. This is not surprising given the observation noted earlier that even though the TMAP and GAP algorithms were different, ALGO patients in each study did better than TAU patients.
Assessments of the importance of the STAR*D trials vary. Insel15 highlighted the importance of the real-world disease management approach of the STAR*D study, its focus on remission rather than response, and the early results showing that patients required higher doses of citalopram for longer periods than previously thought to achieve remission. Menza16 evaluated several STAR*D reports and suggested that the project only incrementally added to our knowledge of how to use antidepressant medications and would probably not have a great impact on how clinicians prescribe these medications. The results do, however, confirm the findings of other studies and the observations of many clinicians about the management of treatment-resistant depression.
It is well worth noting that, despite the comprehensive approach adopted in the STAR*D and GAP phase 3 studies, only half of the patients involved actually achieved remission.6,17 Clearly, psychiatry has much more to learn about how to treat depression. It is also important to note that even in the ALGO groups, some physicians did not strictly follow the algorithm.8,10 If this occurs in a controlled clinical study, what will happen in general psychiatric practice?
As a group, current studies of the algorithm-based management of treatment-resistant depression may say as much about the process of care as its content. Perhaps the best conclusion that can be drawn from the research thus far is that conscientious, organized, and consistent measurement-based care is as important as, if not more important than, the specific medications used to manage the disorder.
Dr Fauman is the author of Negotiating Managed Care and Study Guide to DSM-IV-TR, both recently published by American Psychiatric Publishing, Inc. He is adjunct clinical associate professor of psychiatry at the University of Michigan and medical director of Magellan Behavioral of Michigan in Farmington Hills.
References1. Suppes T, Dennehy EB, Hirschfeld RM, et al. The Texas implementation of medication algorithms: update to the algorithms for treatment of bipolar I disorder. J Clin Psychiatry. 2005;66:870-886.
2. Trivedi MH, Rush AJ, Crismon ML, et al. Clinical results for patients with major depressive disorder in the Texas Medication Algorithm Project. Arch Gen Psychiatry. 2004;61:669-680.
3. Katon W, Russo J, Von Korff M, et al. Long-term effects of a collaborative care intervention in persistently depressed primary care patients. J Gen Intern Med. 2002;17:741-748.
4. Adli M, Rush AJ, Moller HJ, Bauer M. Algorithms for optimizing the treatment of depression: making the right decision at the right time. Pharmacopsychiatry. 2003;36(suppl 3):S222-S229.
5. Rush AJ, Fava M, Wisniewski SR, et al. Sequenced treatment alternatives to relieve depression (STAR*D): rationale and design. Control Clin Trials. 2004;25:119-142.
6. Adli M, Bauer M, Rush AJ. Algorithms and collaborative-care systems for depression: are they effective and why? A systematic review. Biol Psychiatry. 2006;59: 1029-1038.
7. Katon W, Von Korff M, Lin E, et al. Stepped collaborative care for primary care patients with persistent symptoms of depression: a randomized trial. Arch Gen Psychiatry. 1999;56:1109-1115.
8. Adli M, Berghofer A, Linden M, et al. Effectiveness and feasibility of a standard stepwise drug treatment regimen algorithm for inpatients with depressive disorders: results of a 2-year observational algorithm study. J Clin Psychiatry. 2002;63:782-790.
9. Suppes T, Rush AJ, Dennehy EB, et al. Texas Medication Algorithm Project, phase 3 (TMAP-3): clinical results for patients with a history of mania. J Clin Psychiatry. 2003; 64:370-382.
10. Dennehy EB, Suppes T, Rush AJ, et al. Does provider adherence to a treatment guideline change clinical outcomes for patients with bipolar disorder? Results from the Texas Medication Algorithm Project. Psychol Med. 2005;35:1695-1706.
11. Trivedi MH, Rush AJ, Wisniewski SR, et al. Evaluation of outcomes with citalopram for depression using measurement-based care in STAR*D: implications for clinical practice. Am J Psychiatry. 2006;163:28-40.
12. Rush AJ, Trivedi MH, Wisniewski SR, et al. Bupropion-SR, sertraline, or venlafaxine-XR after failure of SSRIs for depression. N Engl J Med. 2006;354:1231-1242.
13. Trivedi MH, Fava M, Wisniewski SR, et al. Medication augmentation after the failure of SSRIs for depression. N Engl J Med. 2006;354:1243-1252.
14. Fava M, Rush AJ, Wisniewski SR, et al. A comparison of mirtazapine and nortriptyline following two consecutive failed medication treatments for depressed outpatients: a STAR*D report. Am J Psychiatry. 2006; 163:1161-1172.
15. Insel TR. Beyond efficacy: the STAR*D trial. Am J Psychiatry. 2006;163:5-7.
16. Menza M. STAR*D: the results begin to roll in. Am J Psychiatry. 2006;163:1123.
17. Rubinow DR. Treatment strategies after SSRI failure-good news and bad news. N Engl J Med. 2006; 354:1305-1307.