Several pharmacodynamically-based—and inter-related—reasons for initiation of APP have been cited in the literature [Figure 3; (1, 8)]. Among these, unsurprisingly, an aim to enhance efficacy and broaden clinical effect into less responsive symptom domains is commonplace, as is the intent to adjust antipsychotic dose vs. adverse event issues (AE). The more direct target-focused reasons include a desire to optimize D2 receptor occupancy, and/or to achieve an overall more favorable treatment response (efficacy and/or AE) outcome by pharmacologically accessing other receptor categories and/or subtypes.

A number of concerns with APP have also been raised (1, 9). Some of the more common actually contrast with the stated intentions for APP. For example, published data suggest that APP often implies higher (instead of lower) total dosages of antipsychotics and increased (rather than attenuated) risks for AE (1, 10, 11)—tentatively related to increased total antipsychotic dosage. With more than one drug on board, there might also be a greater risk of drug-drug interaction events, and difficulties attributing a beneficial or undesired response to the individual antipsychotic agent in an APP treatment combination.

Against the above backdrop it appears reasonable to ask whether APP is

Which antipsychotics are common in APP contexts then? In the literature [e.g., (12, 13)], recurrently emerging drug choices for APP are first- and second-generation (FGA and SGA, respectively) antipsychotic agents like haloperidol (HAL), olanzapine (OLA), risperidone (RIS), quetiapine (QUE), and clozapine (CLZ). In addition, the earliest of the third generation antipsychotics (TGA), aripiprazole (ARI), appears to be a common APP add-on choice to many of the above FGA and SGA class drugs [see, e.g., (14)]. A number of recent case reports likewise suggest that cariprazine (CAR) may emerge as a beneficial option in the APP context [(15–18); vide infra], whereas so far there does not seem to be any reports of the use of the latest TGA brexpiprazole in APP combination approaches. A brief overview of antipsychotic target profiles toward APP is shown in Table 1 below.

A recent review and meta-analysis of the literature (4) found superiority of APP vs. monotherapy in open-label, low-quality studies. However, method-related factors and confounders limited the generalizability of interpretations and conclusions, and in corresponding high-quality, double-blind, randomized studies there was support for enhanced efficacy only for selected APP strategies vs. monotherapy. Specifically, while no superiority was found for combinations of two FGA/SGA D2 antagonists, addition of the partial DA agonist antipsychotic ARI to CLZ medication significantly improved negative symptoms compared to CLZ alone. Given the basic pharmacology profile of CAR it appears reasonable to assume that this agent would work at least equally well to ARI as adjunct to CLZ. Actually, recent case reports concurs with this suggestion [(16, 17); vide infra], although larger, high-quality, double-blind, randomized studies will be needed for further substantiation.

This said, in schizophrenia the identification of factors at the individual patient level will be particularly important to increase the chance of success with an APP approach, thereby promoting personalized treatment in this very heterogeneous patient population. In short: finding the “right drug combination” to the “right patient” is key!

While tolerability and safety concerns may cause some hesitancy to start APP, it should be noted that “…not all antipsychotic combinations are created equal” (8). In addition, it is not the APP approach per se that is the issue, but rather the composition of the specific agents and doses comprised therein that matter. In this regard, a systematic review and meta-analysis of randomized controlled trials comparing APP and monotherapy found no differences regarding intolerability-related treatment discontinuation (4). Moreover, again attesting to the above, it is more likely that APP strategies involving a greater total antipsychotic dose, and thus net target (e.g., D2 receptor) occupancy (8)—at least by antagonist agents [see, e.g., (19)]—would be more liable to increase the AE burden. Conversely, therapeutically useful effects on AE outcome may be achieved by partial DA agonist add-on to CLZ, RIS, OLA, or HAL, to relieve metabolic- and prolactin (PRL)-derived issues (vide infra). Although so far available data are insufficient to allow definitive conclusions it would appear that—contrary to what may be widely presumed—there is a priori no general tolerability, AE, or safety [including mortality; e.g., (20, 21)] reason to discard APP as a possible strategy for a patient in need thereof.

The action of agents in the antipsychotic class must be gauged against their individual pharmacologic target profiles, as many carry multiple receptor affinities and activities.

The dopamine (DA) D2 receptor is a pharmacological target shared by all antipsychotics in current use. However, the affinity for the target varies considerably among drugs in the class—from very high in, e.g., HAL, to pretty poor in, e.g., CLZ and QUE. Moreover, mechanistically the TGA agents ARI and CAR act as partial agonists rather than full antagonists at the D2, D3, and 5 HT1A receptor sites.

In addition to the above, the majority of FGA and SGA also act as blockers of several other neuroreceptor sites with clinical bearing. Table 2 lists key antipsychotic targets and clinically observed outcomes (desired and AE) associated with antagonism or partial agonist drug action at the corresponding sites.

When antipsychotic drug monotherapy responding is an issue—be it for efficacy, AE/tolerability, safety, adherence, and/or other reasons—the treating physician is faced with a number of options and decisions. These may include to adjust the dose, to switch to another antipsychotic agent, and/or to consider augmentation approaches. Irrespective of which tactic is ultimately selected, apart from other clinically-based reflections there are some common basic aspects that require consideration in this situation:

Is the reason for failure insufficient drug exposure (e.g., poor adherence, PK factors)?

Is the reason inadequate efficacy, intolerable AE, and/or safety issues?

If none of the above seem to handle the issue at hand, is APP a worthwhile approach? If so, and similarly to deliberations in a switch situation, thorough attention is recommended to (i) establish the intended therapeutic goal/s (e.g., efficacy domains, AE), (ii) select suitable antipsychotic(s) to use from the PD as well as PK perspective, taking desirable as well as unwanted outcomes into account, and (iii) work out a well-planned strategy to accomplish the goal in mind—with patient buy-in!

In patients for whom APP is deemed to be a fruitful treatment strategy a thorough scrutiny of available options from a pharmacological perspective is advisable. To this end, a selection of APP options is presented below, and examined from a basic PD vs. predicted clinical action perspective. Keep in mind though, that the relative dose/exposure of the selected antipsychotic drugs in an APP combination determines the global response, and hence that the relative clinical benefit/AE outcome in the individual patient may vary.

Refractoriness and persistent prominent negative symptomatology is a frequently mentioned basis for initiation of APP. To this end, administration of a partial DA agonist TGA together with, e.g., CLZ appears to be a particularly appealing option, as it fulfills the criteria of combining two agents with complementary pharmacodynamic (PD) as well as pharmacokinetic (PK) profiles (see, e.g., Figure 5). Indeed, ARI + CLZ is the best documented APP by far, with several studies reporting favorable outcomes both regarding efficacy (not least vs. negative symptoms), tolerability and AE [e.g., (4, 32)].

In addition, Tiihonen (33) recently reported on various APP vs. monotherapy variants with regard to risk for psychiatric rehospitalization in a nationwide Finnish historical cohort (>62,000 patients) of adult schizophrenia. They found that the CLZ + ARI combination conferred the strongest protection against rehospitalization (hazard ratio, HR = 0.42). Interestingly, among the 10 top options (lowest HR) in this regard nine were APP, seven of the APP included CLZ, and the only monotherapy was CLZ—attesting to APP usefulness, as well as to the distinctive position of CLZ in schizophrenia treatment.

Interestingly, as seen in Figure 5, the complementarity in neuroreceptor target profile patterns accomplished with a CLZ + ARI APP may be mimicked to a great extent also by CLZ + CAR, OLA + CAR, and QUE + CAR combinations. From the PD standpoint it would appear reasonable to assume that the potent D2 and D3 (and possibly also 5-HT1A) partial agonist properties of CAR will—similarly to ARI (see, above)—result in a therapeutically advantageous APP via complementation of a relative lack of strong interactions with these targets in CLZ, OLA, and QUE. In fact, CAR may provide a particularly interesting choice, given its prominent D3 affinity and proven clinical effect against primary negative (34) and cognitive (35) symptoms, as well as extended PK half-life (36). Although only limited clinical data with CAR in APP are hitherto available (vide infra), it may be hypothesized that augmentation with this agent might improve efficacy, counterbalance sedative and metabolic AE issues while maintaining the low EPS propensity of CLZ, OLA, and QUE, but also potentially elicit (typically mild, transient) akathisia. It has been suggested that 5-HT2A antagonist may be a valuable option to the β-adrenoceptor blocker propranolol against antipsychotic-induced akathisia (37). Whether or not such a component in CLZ, OLA, and QUE may serve to attenuate any akathisia triggered by CAR however remains to be established.

The RIS + CAR- and HAL + CAR-based APP options are possible, though pharmacologically more complex possibilities (Figure 6). Firstly, the D3 receptor partiality of CAR adds a complementary target effect, presumably advantageous from the negative and cognitive symptomatology viewpoint [e.g., (34, 38, 39)]. However, like CAR both RIS and—in particular—HAL possess appreciable D2 receptor affinities, thereby significantly occupying such sites at therapeutic dosing. In turn, this means that the overall clinical outcome of such combinations with CAR will depend on the relative dose (/concentration) ratio between RIS or HAL and CAR, and thus be more difficult to generalize and forecast. It is possible that the high affinity and partiality of CAR at the D2 (and 5-HT1A) sites may contribute to a lower risk for RIS- and HAL-induced EPS and hyperprolactinemia [see, e.g., (27, 30, 31)]. On the other hand, it is also conceivable that because of their moderate-to-high affinity D2 receptor blockade RIS or HAL may partially counter (even obliterate?) the partial D2 receptor agonism-mediated therapeutic benefits of CAR. The differences in drug half-lives among RIS and HAL vs. CAR (vide infra) may also add to these complexities, thereby contributing to variability in the therapeutic outcome across the 24 h cycle [see, e.g., (14)]. Taken together, it would appear that finding the optimal dosing for these combinations may be challenging, and thus it is likely that a switch from RIS or HAL to CAR would in fact be more preferable.

APP combinations like CLZ + RIS, and OLA + RIS appear pharmacologically less desirable. Thus, whereas the poor D2 affinity of CLZ may be complemented by the higher D2 affinity of RIS, the latter agent (together with its active metabolite paliperidone; PAL) is more liable to cause EPS and hyperprolactinemia. Moreover, as both agents possess significant alpha1-adrenoceptor and H1-receptor antagonism, a CLZ + RIS APP intervention may result in an enhanced acute (orthostatic) hypotensive action as well as risk for accentuated sedation [see, e.g., (40)]. A similar reasoning may apply to the OLA + RIS combination, where any possible benefits of supplemented D2 receptor blockade may be potentially outweighed by an increased AE liability mediated by the very same site (e.g., prolactin rise, increased EPS risk), but also by other targets—as in the CLZ + RIS discussion above. It would appear that within the limitations of available reports (mostly from open, small, short-term, unblinded, non-RCT studies, or case series), clinical outcome data (efficacy and AE) with the aforementioned APP approaches do not generally demonstrate enhanced efficacy but largely concur with the pharmacology-derived reasoning above (41–47).

Among even more questionable (or from a pharmacological perspective, even futile) APP combinations are CLZ + QUE, and ARI + CAR. CLZ and QUE are both rather poor D2 receptor antagonists, and share many other target properties as well (e.g., antagonism of alpha1, muscarinic, H1 sites; see, Figure 6). Thus, the pharmacology-based likely lack of potential for efficacy improvement, together with a possible/probable accentuation of AE liabilities (e.g., sedation, CV, and QTc risks) renders this a pointless APP combination exercise. Only limited clinical data with the CLZ + QUE combination are found in the literature, but appear consistent with the pharmacology-based considerations given (48). The pharmacological profiles of CAR and ARI are by and large overlapping, with the exception that CAR has the decidedly higher D3 receptor affinity of the two [e.g., (22)]. It follows that an ARI + CAR-based APP combination would be futile, whereas when the D3 receptor partiality of CAR is a desired therapeutic property a switch from ARI (to CAR) might be a feasible option. To the best of my knowledge, there are no clinical reports from trials with this latter APP combination.

Taken together, for the reasons discussed above neither of the aforementioned APP combinations (illustrated in Figure 6) are ideal choices from a pharmacological perspective.

Antipsychotics clearly differ in AE liabilities and severity, with TGA being generally more benign than FGA and SGA [e.g., (54)]. This also applies regarding the propensity to induce weight gain and accompanying metabolic AE, with the SGA OLA and CLZ displaying the most, and TGA agents like ARI and CAR the least, harmful profiles (55). Additionally, marked antipsychotic drug heterogeneity in prolactin-raising and sedation-inducing properties occur (40, 56). From the APP perspective it is notable that add-on treatment with TGA can significantly attenuate OLA- and CLZ-induced AE like the aforementioned (57–60).

So far, only very limited data on APP involving CAR is available. However, De Berardis et al. (16) recently reported on CAR add-on in two patients, with comparable illness and medication backgrounds, but only partially responding to CLZ treatment. In both of these cases the CAR addition within 6–8 months brought about a marked remission across symptom domains as scored by PANSS (Figure 7). Interestingly—and in line with the above predictions—body mass index (BMI) dropped an impressive ~3 units from baseline over the same period for both patients. These findings thus support the view that the APP combination of an antagonist and a partial DA agonist antipsychotic agent with complementary PD profiles and short- vs. long-acting PK properties may result in an advantageous outcome both with regard to the desired efficacy and unwanted AE features of the treatment.

Interestingly, a recent single-patient case report suggests that CAR plus QUE may also be an attractive APP option (18). The combined treatment with these two antipsychotics resulted in successful alleviation of cognitive and negative impairments in a young male patient, whom previously had been through a variety of but partially efficacious FGA and SGA regimens. Intriguingly, and in line with other recent case studies, the CAR + QUE APP was associated with an abrupt cessation of smoking, and curbed use of recreational drugs, tentatively indicative of anti-craving effects. In support of this speculation, combined CAR + QUE treatment was reported to markedly attenuate alcohol craving and bring about lasting symptom stability in a bipolar I patient (61). Further, CAR monotherapy has also been reported to result in abrupt remission from persistent methamphetamine psychosis and improved positive as well as negative symptomatology in a treatment-naïve male patient (62), and to benefit two cases of schizophrenic patients with other substance use disorders (63). These clinical observations are consistent with preclinical literature and theory suggesting that D3 receptors may be implicated in drug dependence issues (64–66).

What would be the potential PD mechanistic substrate/s underlying partial agonist-induced improvements of core symptoms and AE when incorporated into an APP approach? From a theoretical viewpoint it appears plausible that the effects on positive and negative symptomatology involve partial agonism at the DA D2 and D3 receptor sites (Table 5). A direct D2 receptor interaction is also highly likely to explain the normalization of antagonist-induced hyperprolactinemia. However, the reported beneficial effects on anthropometric and metabolic issues, as well as the offsetting of sedation is hypothesized to be the result of counterbalancing neuronal circuits rather than direct competition between drugs at a particular target (Table 5). That is, the blockade of, i.a., H1 and 5-HT2C receptors in some brain regions/neurocircuitries drives the weight, metabolic parameters, and sleepiness in one direction, whereas partial D2 agonism drives them the opposite way. Indeed, generally, agents in the TGA partial DA agonist class are considered less metabolically adverse [see, e.g., (55)], and also more “activating” and less “sedating” as compared to FGA and SGA (67).

In patients for whom APP is deemed to be a worthwhile treatment strategy, it is evident from the above that a thorough scrutiny of available options is advisable. By and large, whereas combinations of (FGA and SGA) D2 receptor antagonists may be challenging, available data discussed in this account indicate that from the pharmacological perspective selected APP, in particular based on SGA + TGA, may indeed be efficacious, tolerable, safe, as well as useful in a preventive, relapse/re-hospitalization context.

Within this APP framework, a PD comparison between the TGA:s CAR and ARI suggests that while both display high affinity partial agonist activity at the D2 receptors, CAR displays even higher affinity for the D3 than the D2 sites and is nearly 10-fold more potent than ARI at D3 receptors [e.g., (22)]. It may be hypothesized that, although similar in their efficacy against positive symptoms the appreciably stronger D3 action of CAR vs. ARI may translate to an improved profile toward primary negative symptoms—and, speculatively, also when dependence issues may be involved (see, above). Further, while both agents may elicit mild and typically transient akathisia, neither appears burdened by marked EPS, metabolic issues, prolactin rises, or sedation (68). As both agents have long t_1/2 (CAR > ARI; see, Table 3), an SGA + TGA combination in an APP strategy setting thus also fulfills the PK aim to match a short-acting with a long-acting agent.

From an efficacy point-of-view it appears probable that APP will be prescribed for (i) patients with predominant and persistent negative symptoms, and (ii) patients with residual positive symptoms; e.g., patients with chronic auditory verbal hallucinations, only partly alleviated by antipsychotic monotherapies. Unfortunately, available literature does not seem to shed very much light on whether a particular type of APP would be preferred for one vs. the other of these forms of enduring issues. However, as a recommended Guideline sequel to ≥ 2 failed monotherapy trials, it appears logical that CLZ would be highly prevalent in any strategies to deal with persistent residual symptoms—irrespective of domain. While agents with high affinity and antagonism or partial agonism at the D2 receptors are commonly used to attenuate positive symptoms the negative/cognitive domains appear overall less susceptible and more difficult to reach, even with CLZ. From a pharmacological standpoint partial agonist TGAs like ARI or CAR should be useful to boost efficacy of CLZ, although so far high-quality data are supportive only for ARI with respect to negative symptoms [discussed above, see (4)]. The notable efficacy of CAR monotherapy in this latter indication (34, 35) may possibly suggest a further edge of this agent in APP when negative/cognitive issues dominate the clinical picture. Until further studies to assess this prediction, it however must remain purely speculative.