PROSPECT (Prevention of Suicide in Primary Care Elderly—Collaborative Trial) aims at decreasing suicides in older primary care patients by improving the recognition and treatment of late-life depression (+Bruce and Pearson, 1999). Several studies have documented that even when primary care physicians (PCPs) diagnose depression in one of their patients, they often do not treat it or treat it inadequately (+Schulberg et al., 1997). However, some recent studies have demonstrated that PCPs can be trained to manage depression adequately (+Schulberg et al., 1996; +Wells et al., 2000). This observation suggests that the inadequate management of depression in primary care settings is probably due to structural issues rather than deficient knowledge or lack of fundamental skills. Callahan and others have argued that competing demands (e.g. need to address multiple physical problems) impede the PCPs’ ability to manage psychosocial problems and depression in their patients (+Callahan et al., 1996; +Rost et al., 2000). Katon and his colleagues have shown that the treatment of depression in a large primary care clinic can be improved by the implementation of a ‘collaborative model’ according to which patients are managed by their PCP in close collaboration with a psychiatrist (+Katon et al., 1992, +1995, +1999). Despite its success in several research studies, this model has not been implemented in clinical practice. Thus, as described elsewhere in this volume (+Schulberg et al., 2001), PROSPECT is testing a different model. In PROSPECT, a specially trained master-level clinician (the ‘health specialist’) works in close collaboration with a depressed patient’s PCP to implement a comprehensive disease management program. When a patient had been diagnosed with a depressive syndrome that requires treatment (see +Raue et al. (2001) elsewhere in this volume), PROSPECT health specialists implement the various clinical tasks necessary for a successful treatment outcome, including educating older depressed patients and their family about depression, identifying and addressing comorbid physical and psychiatric conditions interfering with antidepressant treatment, monitoring adherence, managing treatment-emergent adverse effects and regularly assessing change in depressive symptoms to evaluate whether the current treatment is effective or whether it needs to be modified (+Schulberg et al., 2001). A depression treatment manual has been developed to assist PROSPECT health specialists. This manual includes an algorithm guiding the selection of antidepressant medications. This paper presents this algorithm and describes the rationale underlying its design.

Before developing a specific treatment algorithm for PROSPECT, three major guidelines for the treatment of depression were considered (+American Psychiatric Association, 1993; +Depression Guideline Panel, 1993; Gilbert et al., 1999; +Rush et al., 1999). In 1993, the Agency for Health Care Policy and Research (AHCPR) published evidence-based guidelines for the diagnosis and treatment of depression in primary care (+Depression Guideline Panel, 1993). The general principles proposed by the AHCPR guidelines (e.g. the goal of acute antidepressant treatment should be full remission of symptoms rather than symptomatic improvement; depressed patients need to receive continuation treatment following acute response) have been incorporated in the treatment approach for PROSPECT patients. However, because few empirical data on the effectiveness of newer antidepressants had been published when the AHCPR guidelines were developed, these guidelines focus on the use of tricyclic antidepressants (TCAs) that are no longer considered optimal, in particular in older patients treated by PCPs. This is also the case for the guidelines published by the American Psychiatric Association (APA) in 1993 (+American Psychiatric Association, 1993). In addition, the APA guidelines do not focus on the primary care setting. More recently, an algorithm and guidelines were developed by a group of experts for the treatment of depressed patients utilizing the services of the Texas Department of Mental Health (Gilbert, 1999; +Rush, 1999). This algorithm incorporates the significant changes that have occurred in the pharmacotherapy of depression during the past decade. However, like the APA guidelines, this algorithm was not developed specifically to address the constraints and impediments encountered when treating older primary care patients. Thus, the PROSPECT investigators have developed an algorithm that builds upon these existing guidelines after updating them and adapting them to the special circumstances of older primary care patients. Ideally, treatment guidelines would be evidence-based. However, there are few published data directly relevant to the pharmacological treatment of geriatric depression in primary care settings. Thus, treatment parameters in the PROSPECT treatment algorithm (e.g. selection of a specific agent, target dosage, titration, or duration of treatment) are based on other relevant data or on expert consensus opinion.

Assuming that an agent with demonstrated efficacy is used, dose and duration of treatment are the major determinants of the success or failure of an antidepressant trial. In turn, the side effect profile of a given agent often determines the likelihood that an appropriate dose would be taken for an appropriate duration (+Mulsant and Pollock, 1998). Most randomized clinical trials conducted under double-blind conditions have failed to demonstrate clear differences in efficacy or tolerability among antidepressants of various classes (+Cohn et al., 1990; +Hutchinson, 1991; +Kyle et al., 1998). However, several meta-analyses have found that adverse effects were more likely to lead to discontinuation in subjects treated with TCAs than in those treated with selective serotonin reuptake inhibitors (SSRIs) (+Montgomery et al., 1994). Similarly, a review focusing on geriatric trials has concluded that discontinuation rates in randomized trials are about one-third lower with SSRIs than with TCAs (+Schneider, 1996). Under typical clinical conditions, patients started on an SSRI are more likely to receive a minimally adequate trial than those started on a TCA or other antidepressants (e.g. trazodone or bupropion) (+Katzelnick et al., 1995). Furthermore, SSRIs have a wide therapeutic index and lack significant anticholinergic or cardiac toxicity (+Reynolds et al., 1997; +Roose et al., 1998). Thus, due to their efficacy, overall safety and ease of use, SSRIs have become first-line antidepressants in both primary care and psychiatric settings (+Mulsant et al., 1997; +Reynolds et al., 1997). Thus PROSPECT recommends the use of an SSRI in all patients who have not yet received any antidepressant or who have been inadequately treated (see below). In the absence of strong evidence suggesting the definite superiority of one of the five SSRIs currently available in the USA (i.e. citalopram, fluoxetine, fluvoxamine, paroxetine and sertraline) (+Kasper and Heiden, 1995), citalopram was selected as the agent of choice because of the limited drug-drug interactions, absence of discontinuation syndrome and limited central nervous system activation associated with its use. Furthermore, some data suggest that it is effective and well tolerated by older depressed patients with and without cognitive impairment (+Gottfries and Karlsson, 1992; +Kyle et al., 1998).

PROSPECT recommends dosages (or blood concentrations for TCAs when they are used) based on published controlled trials, recommendations from drug manufacturers (as presented in ‘package inserts’) and clinical experience. Target dosages are the dosages reported to be efficacious in older patients. In the absence of data for geriatric patients, the recommended dosages are based on the dosages used in younger patients, modified as necessary. Similarly, maximum dosages are extrapolated from maximum dosages used in younger patients, taking into account the pharmacokinetic profile of the specific agent and changes in drug metabolism associated with aging (see Table 1+). For instance, citalopram treatment is initiated as follows: 10 mg is given at bedtime on the first day, 20 mg for the next 6 days, and 30 mg from the eighth day on. Based on tolerability, the PCP and the health specialist can modify the titration and the target dosage.

Premature discontinuation of acute antidepressant treatment in primary care and psychiatric practice has been associated with low response rates (+Bonner and Howard, 1995; +Amsterdam and Hornig-Rohan, 1996; +Mulsant and Pollock, 1998). However, there is no consensus on how long a clinical antidepressant trial should last, particularly when patients fail to demonstrate significant improvement in their depressive symptoms. Published antidepressant trials in older depressed patients have typically lasted only 4—6 weeks (e.g. +Tollefson et al., 1995). Longer trials have shown a higher response rate as the length of the trial increased (+Greenhouse et al., 1987; +Georgotas and McCue, 1989; +Little et al., 1998). In one study in which depressive symptoms were assessed weekly over 16 weeks, though some older patients experienced a rapid resolution of their depressive symptoms (i.e. within 2—4 weeks), improvement was delayed for up to 10 weeks in 40% of the patients who eventually demonstrated a sustained response (+Dew et al., 1997). A re-analysis of the databases of two comparable antidepressant trials conducted in 214 midlife and in 148 older patients found that the overall remission and recovery rates were similar in the two age groups but, on average, older patients experienced a slower reduction of their depressive symptoms (+Reynolds et al., 1996). Patients with dysthymia or comorbid anxiety disorders, who are seen frequently in primary care settings, have also been shown to respond to antidepressant treatment but to require longer trials (+Mulsant et al., 1996; +Thase et al., 1996). Thus, in PROSPECT, the full duration of the acute treatment phase is set to last ≥12 weeks and antidepressant trials started prior to enrollment are considered inadequate if they have been shorter than 8 weeks.

As described elsewhere in this volume (+Raue et al., 2001), when the PCP and the health specialist implement a new antidepressant trial, the health specialist monitors depressive symptoms longitudinally with the 24-item Hamilton Depression Rating Scale (HDRS). After a patient has been treated for 6 weeks, the target dosage is maintained if the patient demonstrates at least a 50% reduction in depressive symptoms on the HDRS. If a patient has a partial response (i.e. a 30—50% reduction in HDRS score), the dosage is increased (see Table 1+). If the patient has an insignificant response (i.e. a reduction in HDRS score of <30%), the treatment is discontinued. After 12 weeks, treatment response is re-evaluated and the following actions are taken: if a patient has achieved symptomatic remission (i.e. an HDRS score of ≤10), continuation treatment is initiated at the dose that led to remission. Patients who remain symptomatic (i.e. with an HDRS score of >10) but who have experienced a significant reduction in depressive symptoms (i.e. reduction in HDRS score of ≥50%) are considered partial responders and guidelines for augmentation are followed. Table 2+ lists the factors that favor an augmentation rather than a substitution strategy (i.e. combining two antidepressants rather than switching to another antidepressant). Based on these factors, in partial responders, the algorithm favors first a dosage increase up to the maximum recommended dosage followed by augmentation of current treatment. Finally, patients who remain symptomatic (i.e. those with an HDRS score of >10) and who have not experienced a significant reduc tion in depressive symptoms (i.e. reduction in HDRS score of <50%) are considered non-responders and guidelines for switching to another antidepressant are followed. Thus, in PROSPECT, unless a patient experiences severe adverse effects, no trial lasts <6 weeks and most trials last ≥12 weeks. Also, to be considered non-responders, patients have to fail to demonstrate either minimal improvement after 6 weeks of treatment at target dosage or significant improvement after dosage has been increased to the maximum recommended dosage and 12 weeks of treatment have been completed.

Older depressed patients often report somatic symptoms such as dizziness, tremulousness, sweating, headache, tinnitus, diarrhea, stomach complaints and palpitations (+Koenig and Blazer, 1992). Successful treatment of depression is often associated with a concomitant decrease in somatic complaints (+Reimherr et al., 1990; +Miller et al., 1991; +Rollman et al., 1997). However, early in treatment, patients, their family or clinicians may mistake these somatic symptoms for side effects. Unfortunately, this often leads to premature discontinuation of antidepressant medication (+Bonner and Howard, 1995; +Mulsant and Pollock, 1998). In PROSPECT, the health specialists assess somatic symptoms at baseline prior to initiation of treatment and during follow-up (longitudinally) (+Raue et al., 2001). This helps to determine whether a somatic complaint is new and may indeed represent a side effect or whether it was already present prior to initiation of treatment. Patients (and their family) are informed by their health specialists about the possible occurrence of specific side effects and advised to contact them if significant side effects occur. When side effects occur, health specialists provide support and, if warranted, they ask the PCPs to adjust the dosages, time of administration (e.g. switching a medication from AM (morning) to HS (bed time) in case of a complaint of daytime sedation), or to institute symptomatic treatment (e.g. stool softeners). Thus, in many cases, patients’ expectations that they may feel temporarily worse, support from the health specialist and slight medication adjustments forestall premature discontinuation of antidepressant therapy. Nonetheless, some older depressed patients develop significant side effects that they are unable to tolerate and their medications have to be discontinued. In PROSPECT, patients who are unable to tolerate even citalopram 10 mg/day are treated with bupropion SR. If they are also unable to tolerate bupropion SR, they are switched to nefazodone. Citalopram, bupropion and nefazodone affect different neurotransmitter systems and have distinct side effect profiles. Thus, it is unlikely that a patient would not be able to take any of these medications. However, when patients seem unable to tolerate any antidepressant (or when they prefer to avoid pharmacotherapy), health specialists can provide interpersonal psychotherapy (IPT).

As discussed above, many older patients who appear to have failed to respond to an antidepressant have not actually completed an adequate trial. Thus, in PROSPECT, when new subjects, their family or even their PCP report that they have failed to respond or to tolerate one or more antidepressants in the past, a new trial is instituted with a PROSPECT first-line antidepressant (i.e. citalopram). By paying attention to doses, management of side effects, adherence and duration of treatment, many of these apparently ‘treatment-resistant’ patients demonstrate a satisfactory response to a well-conducted trial. Similarly, when a new subject is already being treated with an antidepressant but he or she remains symptomatic, the health specialist first optimizes the current trial (see Table 3+). This ensures that depressed patients have truly failed to respond to the maximum recommended dose given for an adequate duration before they are declared partial or non-responders. When patients fail an adequate trial, i.e. when they demonstrate treatment resistance, a systematic approach is implemented (see Table 3+).

The general principles that have guided the ‘ranking’ of various therapeutic options are as follows. Based on the potential advantages and risks (or costs) of an augmentation strategy compared with a substitution strategy (see Table 2+), augmentation of the current agent is preferred in partial responders and switching antidepressant is favored in non-responders. Of note, IPT may also be combined with antidepressants. As discussed above, IPT alone may also be used as an alternative to pharmacotherapy. The use of IPT as a substitute vs augmentation depends on the tolerability of the medication and the presence or absence of partial response. For instance, in patients with milder depression, who do not seem to tolerate antidepressants, IPT alone may be a preferred treatment since psychotherapy has been shown to have comparable efficacy to pharmacotherapy in the treatment of milder depression (+Schulberg et al., 1996; +Thase et al., 1997). However, in patients with more severe depression or those who show only partial response to pharmacotherapy, combination treatment may be preferred since combined psychotherapy and pharmacotherapy has been shown to be superior for the treatment of more severely depressed patients (+Miller et al., 1989; +Bowers, 1990; +Thase et al., 1997).

Treatments for which there is empirical evidence are favored. For instance, venlafaxine is favored over mirtazapine in treatment-resistant patients (+Nierenberg et al., 1994; +de Montigny et al., 1999). However, treatments that are often poorly tolerated by elderly patients are given lower priority than treatments more likely to be tolerated, even when the efficacy of the latter treatments may be less well established. For instance, bupropion augmentation of SSRIs is favored over lithium augmentation, although more research is available on lithium augmentation. Also, despite insufficient empirical evidence, this treatment is favored because pharmacodynamic interactions suggest a possible synergistic effect (as opposed, for instance, to a combination of two SSRIs).

Drug trials that are simpler to implement in primary care are favored over drugs of known efficacy that require special procedures. For instance, nortriptyline is given low priority because it requires serial ECGs and monitoring of plasma concentrations, it has significant anticholinergic side effects and it is unsafe in overdose. Similarly, monoamine oxidase inhibitors (MAOIs) are not favored because they require a special diet and can be involved in lethal drug-drug interactions.

The PROSPECT treatment algorithm emphasizes the use of pharmacotherapy because medication management is often more practical in primary care settings. Thus, this emphasis will make the results of the study more generalizable and more likely to influence clinical practice. Thus, formal (i.e. interpersonal) psychotherapy is reserved for patients who do not tolerate or do not respond to pharmacotherapy. However, the psychoeducation, monitoring and support provided to all patients and their family by the PROSPECT health specialists are expected to play a crucial role in the success of the intervention. The PROSPECT investigators have hypothesized that the implementation by a health specialist working closely with a PCP of the systematic approach captured in this algorithm will significantly improve adherence and outcomes. The design of PROSPECT permits this hypothesis to be tested (+Bruce and Pearson, 1999). It also allows an empirical validation of the proposed treatment algorithm based on the proportion of older depressed primary care patients who respond to ‘first-step treatment’ or who require more intensive interventions. Finally, this paper focuses on the description of and rationale for the proposed sequential approach to the pharmacological treatment of older depressed patients participating in PROSPECT. However, PROSPECT implements a broader depression management program. In addition to the treatment algorithm, the treatment manual contains strategies developed to handle a broader array of factors involved in treatment non-response such as diagnostic heterogeneity, comorbid conditions (e.g. alcoholism or cognitive impairment) and personality. In the future, the contribution to positive outcomes of these interventions, above and beyond adherence to treatment guidelines, will also need to be evaluated.