Sequencing treatments requires initially selecting a single-modality treatment. Given that acute treatment with CBT or antidepressant medication is equally likely to be effective for most patients (
22,
59), the choice of treatment may be driven by patient preference and other practical factors. The specific form of any added second treatment will depend on the patient status at the end of monotherapy. Specifically, the second treatment can be designed to address a lack of response to the initial treatment (i.e., patient still in a full major depressive episode), residual symptoms after improvement with initial treatment, or prevention of relapse and recurrence.
Despite the large number of studies examining sequential treatment of depression, the variety of designs used limits clear interpretation of the results. Some trials have compared an active second intervention, typically psychotherapy, with treatment as usual, in which patients may or may not adjust their mental health treatments on the basis of their preferences and clinician’s recommendations. Treatment-as-usual designs are relatively easy to implement, but treatment as usual is a weak comparator arm for sequential treatment combination studies because a significant percentage of patients may not be receiving an active treatment of any kind (
38,
60). Moreover, treatment-as-usual comparisons carry a significant bias in favor of finding efficacy of the added treatment because of placebo responses. Specifically, because blinding of patients in such designs is not possible, patients randomly assigned to treatment as usual may experience demoralization effects of not getting the desired treatment, whereas those in the active condition have greater interaction with mental health professionals, with consequent mobilization of known placebo effects. In medication-augmentation trials, improvement after open-label addition of a second medication is not considered adequate evidence for a specific beneficial effect of the medication because improvement may derive from nonspecific benefits of attention, support, and mobilization of hope (
61). However, it is also possible for treatment-as-usual designs to bias results toward finding no benefit from an additional treatment because in both groups other treatments are not controlled, modifications of which may have greater effects on outcomes than the active intervention. Stronger sequential treatment designs involve randomization to two intervention arms in which patients can have reasonably similar expectations of gaining benefits.
Another important caveat for results of sequential treatment studies is their risk for bias arising from participant selection factors. Specifically, when all trial participants receive a specific initial treatment modality (whether medication or psychotherapy), the trial may experience selection bias, enrolling patients who are specifically seeking the initial treatment modality (or, alternatively, who strongly dislike an alternative). Another limitation of sequential treatment strategies is that patients may refuse to enter the second (sequential) phase of treatment. In a study of 141 patients with mild to moderate major depressive disorder who were initially randomly assigned to short-term supportive psychodynamic therapy or antidepressant treatment, those with <30% improvement (by the Hamilton Depression Rating Scale) by the eighth week of treatment were offered the alternative treatment (
62). Of the 63.6% of the patients who showed inadequate improvement, nearly 40% of the patients refused the additional treatment (
62). Similarly, in the Sequenced Treatment Alternatives to Relieve Depression (STAR*D) study, 71% of patients not in remission with citalopram monotherapy were not willing to be considered for possible randomization to CBT (
63). Reasons for patients being unwilling to receive additional treatment are unknown but may include time burdens, demoralization from lack of improvement with initial treatment, or sufficient satisfaction with partial improvement that reduces motivation to pursue additional treatment to achieve full remission.
Sequential Combination After Nonresponse to Monotherapy
Most studies evaluating sequential combination treatment among patients failing to respond to a monotherapy have tested the addition of psychotherapy after antidepressant medication. It is surprising that so few studies have examined addition of an antidepressant after poor response to psychotherapy, given that most patients prefer psychotherapy to antidepressant medication (
64). Furthermore, many patients in clinical care initally receive psychotherapy, albeit often of limited duration or supportive rather than evidence-based (
65,
66). The potential value of sequential addition of an antidepressant for poor responders to psychotherapy is suggested from a study of women with recurrent major depressive disorder treated for up to 24 weeks with interpersonal therapy. Among 86 women who received interpersonal therapy alone and who either failed to respond by week 12 or failed to remit by week 24, addition of an open-label SSRI to ongoing interpersonal therapy produced remission in 67% of the women (
67). Unfortunately, no large randomized trials have compared addition of an antidepressant versus placebo among psychotherapy nonresponders, which would be necessary to quantify the specific benefit of antidepressants in this form of sequential combination treatment.
The largest randomized trial to examine the value of sequential treatments was REVAMP, in which 808 patients with chronic forms of major depression all received an initial 12-week course in phase 1 of algorithm-guided medication management. Patients not in remission by week 12 (N=491) were randomly assigned in phase 2 to another 12 weeks of treatment in one of three arms: antidepressant medication with further algorithm-guided adjustments, continued phase 1 medication plus 16 sessions of CBASP, or continued phase 1 medication plus 16 sessions of brief supportive psychotherapy (
37). The three treatments did not significantly differ in efficacy, achieving similar remission rates (medication only, 40%; CBASP, 39%; brief supportive therapy, 31%). The degree of improvement from phase 1 medication (partial response vs. nonresponse) did not moderate the degree of benefit derived from each of the three treatments (
37). A similar lack of difference between adding cognitive therapy versus augmenting with a second medication was found in the STAR*D trial, in which people not in remission who completed 12–14 weeks of citalopram in step 1 were randomly assigned to second-step treatments. Patients in step 2 who received cognitive therapy added to citalopram had nonsignificantly lower remission rates than the patients who received buspirone or bupropion augmentation of citalopram (23% vs. 33%, respectively; p=.19) (
68).
In contrast, studies evaluating psychotherapy added to treatment as usual have generally found benefits for combination treatment. The CoBalT study enrolled 469 routine clinical care patients who continued to meet full major depressive episode criteria despite having received an adequate course of antidepressant medication treatment (
69). Assignment to 12 sessions of CBT in addition to treatment as usual provided better short-term and long-term outcomes compared with treatment as usual alone. Patients receiving CBT had a response rate (on the basis of a 50% score reduction on the self-rated Beck Depression Inventory–II [BDI-II];
70) more than double that of the treatment as usual group after six months (46% vs. 22%, respectively; p<.001). Self-reported remission rates (BDI-II<10) were also superior in the CBT group (28% vs. 15%, respectively; p<.001). At one-year posttreatment follow-up, remission rates continued to improve in the CBT group but did so only minimally in the treatment-as-usual group (40% vs. 18%, respectively; p<.001) (
69). In a smaller trial of 106 patients with chronic depression, addition of eight 2.5-hour sessions of group-based CBASP to treatment as usual produced superior remission rates to treatment as usual alone (26% vs. 6%, respectively; p=.02) (
70). Another trial found that MBCT combined with treatment as usual was not statistically superior to treatment as usual for remission (17% vs. 6%, respectively; p=.15), and was inferior to combining CBASP with treatment as usual on overall depressive symptoms (
71). Finally, a randomized trial of individually delivered CBASP (provided as 24 sessions over one year) added sequentially to treatment as usual found only small benefits over treatment as usual alone among 139 clinical care adults with chronic forms of major depressive disorder (
38). Although the reduction in depressive symptoms at week 52 was significantly greater in the CBASP group, remission rates were low and not statistically superior to the treatment-as-usual group (19% vs. 10%, respectively; p=.11) (
38).
A sequential treatment strategy that is not truly a form of combination treatment is switching to, instead of adding, a second form of treatment. A patient’s ability to respond specifically to psychotherapy or medication may depend on his or her pretreatment brain activity state (
23,
24), which implies that failure to benefit from one treatment modality should lead to strong consideration for initiating the alternative form of treatment. In this approach, learning obtained from initial psychotherapy may persist and continue to be applied by patients during subsequent treatment, but medication effects are not believed to be sustained. In the study combining nefazodone with CBASP described above (
29), 140 nonresponders to an initial 12-week monotherapy treatment with nefazodone or CBASP were switched to receive the alternative treatment. Remission rates after 12 weeks with the second treatment were numerically, but not statistically, superior among patients switched from nefazodone to CBASP than vice versa (36% vs. 27%, respectively; p=.11), demonstrating that maintenance of the initial treatment is not always necessary after nonresponse, and therefore switching, as opposed to combining, can be a rational strategy (
33). In step 2 of STAR*D, patients not in remission after 12–14 weeks of citalopram who were switched to CBT achieved a 31% remission rate; this outcome did not significantly differ from the 27% rate achieved among those who switched to another antidepressant (
68).
Sequential Combination to Address Residual Symptoms After Monotherapy
The importance of residual symptoms after acute treatment as a potent predictor of eventual return of full-syndrome major depressive episode is one of the most robust findings in depression research (
72–
74). Patients who improve after acute treatment but who have persisting subthreshold symptoms of major depressive disorder carry a significant risk of relapse and recurrence, whether the initial treatment was medication (
31,
75) or CBT (
32,
76,
77). These data have led to great interest in applying combination treatments to achieve full remission (
1). The long-term protective effect of combination treatments has proven to be greatest among those patients with the highest risk of recurrence, with the two most important risk factors being the level of residual symptoms after acute or continuation phase treatment and the number of prior major depressive episodes.
Most trials targeting residual symptoms have added psychotherapy to patients previously treated with an antidepressant medication. An early randomized study found that among 40 patients who improved but had residual symptoms after three to five months of medication, the addition of ten sessions of CBT administered over 20 weeks reduced residual symptoms significantly more than a similar number of sessions of CM (
75). After the ten sessions, all patients were tapered off their antidepressant. During subsequent follow-up, patients who had received CBT had lower rates of recurrence, although by six years, 50% of the CBT group and 75% of the CM group had experienced recurrence (
78,
79). These results were supported by subsequent larger trials. Sixteen sessions of cognitive therapy combined with five monthly sessions of CM delivered over 20 weeks was superior to CM alone among patients with residual symptoms despite at least eight weeks of antidepressant medication (
80). Though relatively few patients in this trial achieved remission after 20 weeks (cognitive therapy and CM, 24%; CM alone, 11%; p=.03), the cognitive therapy combined with CM group experienced a significantly lower rate of relapse-recurrence than the CM group during the one-year follow-up (29% vs. 47%, respectively; p=.02), despite all patients being maintained on an antidepressant through this phase (
80). In contrast to these results, phase 1 partial responders to 12 weeks of antidepressant medication in the large REVAMP trial showed no meaningful benefits in phase 2 from the addition of CBASP or brief supportive therapy to continued medication compared with 12 weeks of algorithm-guided medication alone (
37). The REVAMP trial differed from other sequential trials in that enrollment was limited to patients with chronic forms of major depressive disorder, and it included an active psychotherapy as a comparator.
Two large trials have also found that combining MBCT with treatment as usual for patients with residual symptoms of major depressive disorder reduced levels of depressive symptoms compared with treatment as usual alone (
81,
82) and that these gains were still present after 12-month follow-up (
82). However, in both trials, half or fewer of the patients were taking an antidepressant medication, limiting interpretation of the value of MBCT as a combination treatment for residual depressive symptoms.
Although adding antidepressant medication to address residual symptoms after psychotherapy may appear to be a poorly supported treatment on the basis of analyses finding minimal antidepressant efficacy over placebo among patients who were mildly ill in clinical trials (
83,
84), other analyses have suggested that antidepressants are effective for mild depressive symptoms (
85,
86). Unfortunately, the efficacy of sequential addition of medication after nonremission to psychotherapy has received little empirical evaluation, but limited data suggest this form of combination treatment may have value. In a unique trial design, patients who achieved an unstable remission and who were considered to carry a significantly higher risk for relapse after 16–20 sessions of cognitive therapy were randomly assigned to (a) eight months of continuation cognitive therapy (ten 1-hour cognitive therapy sessions targeting relapse prevention over eight months), (b) initiation of fluoxetine 40 mg/day, or (c) placebo (
87). Remission rates at the end of this continuation phase were nearly identical in the continuation cognitive therapy and fluoxetine groups, and both were higher than placebo (72%, 71%, and 58%, respectively; p=.13) (
87). In another large trial, a similarly high remission rate of 67% was observed among women who did not remit with up to 24 weeks of interpersonal therapy who subsequently then received addition of an SSRI to ongoing interpersonal therapy sessions. Among those who remitted after 17 weeks of continuation treatment with interpersonal therapy plus SSRI, the SSRI was discontinued with a taper, and patients were followed for two years while receiving maintenance interpersonal therapy sessions (
67). Recurrence rates among these patients were significantly higher during follow-up than among those who were in remission with interpersonal therapy alone (SSRI plus interpersonal therapy with subsequent SSRI tapering: 50%; interpersonal therapy alone: 26%) (
67). These data again point to the need to sustain maintenance treatments among patients at high risk of recurrence.