On the Importance of Identifying Mechanisms and Active Ingredients of Psychological Treatments

Abstract

This commentary addresses the thought-provoking article by Lorenzo-Luaces, 2023. We review areas of both agreement and disagreement with the author’s points, noting that readers should not infer that research into active ingredients and mechanisms is pointless. We conclude with a call for more research into the mechanisms of therapeutic change and the active ingredients of therapeutic interventions, with the aim of disseminating treatments that are both effective and efficient.

In his commentary, Identifying active ingredients in cognitive-behavioral therapies: What if we didn’t?, Lorenzo-Luaces, 2023 makes several important points with which we believe merit further discussion. Specifically, he argues: 1. Treatment studies are “messy” settings in which to isolate active processes of interventions; 2. There is a long road from “identifying active ingredients” to improving outcomes; and 3. There are research questions that are closer to improving outcomes. In this brief article, we will review Lorenzo-Luaces’s points and identify areas of agreement and disagreement. We conclude the article with a call for more (not less) research into the mechanisms of therapeutic change and the active ingredients of therapeutic interventions.

Addressing Lorenzo-Luaces’s points

Lorenzo-Luaces’s first point is that treatment studies are “messy” settings in which to isolate active processes of interventions. This is a valid argument. In routine clinical settings, issues such as diagnostic heterogeneity, diagnostic comorbidity, non-standardization of treatment procedures, and variability in mechanism and outcome assessment all contribute to the “mess,” making it difficult to answer critical research questions. That said, there have been numerous calls to enhance the understanding of treatment elements and components by redesigning clinical trials to provide a much clearer focus on testing how treatment elements are linked to mechanism(s) as a prelude to evaluating clinical significance (e.g., Doss, 2004; Nielsen et al., 2018). In addition to trial designs giving primacy to the elucidation of mechanisms, there is a role for dismantling studies, utilization of technology-delivered interventions that can systematize the type and level of exposure to discrete interventions, and intervention optimization methods (e.g., multiphase optimization strategy (MOST) and the sequential multiple assignment randomized trial (SMART) designs; Collins et al., 2007) to aid the identification of active treatment elements and their mechanism of action (Magill et al., 2020), though we do note that such trials often require large samples for adequate statistical power. Indeed, Lorenzo-Luaces makes a similar point, writing, “Indeed, one of the promises of identifying active ingredients is that it will lead to treatments that can be stream-lined and are thus easier to disseminate than existing CBTs (Watkins & Newbold, 2020). I find this among the most compelling arguments for active ingredients work” (p. 4), and we do not want this central point to be lost in the discussion.

Relative to these points, Lorenzo-Luaces makes a valid argument that many putative studies of “mechanisms” have failed to establish temporal precedence, and that therefore it is difficult to understand whether the purported mechanism influences the clinical outcome, whether the clinical outcome influences the mechanism, or whether a third (perhaps unmeasured) variable is driving both of these. We note here that the study of mechanism is distinct from the study of active ingredients (a distinction also made by Lorenzo-Luaces, German, & DeRubeis, 2015), although we also note that the literature has yielded conflicting definitions (see Cohen et al., 2023). For the sake of clarity, here we will use the term “mechanisms” to mean the intrapersonal processes that must change in order to achieve a clinical response (e.g., Koch & Cratsley, 2020). One example of a mechanism is anxiety sensitivity, or a tendency to perceive one’s own physiological and emotional arousal as dangerous (Reiss et al., 1986). Anxiety sensitivity can be found across the anxiety-related disorders, although it appears to be particularly concentrated in panic disorder (Olatunji & Wolitzky-Taylor, 2009). When patients with panic disorder receive cognitive-behavioral therapy, reductions in anxiety sensitivity both precede and predict reductions in panic severity (Gallagher et al., 2013), establishing anxiety sensitivity as a likely mechanism of action. Once identified, that mechanism can then be targeted for engagement by a range of strategies in prevention or intervention models (e.g., Fitzgerald et al., 2021).

Illustration of a mechanism does not, however, inform a clinician about what to do with a patient who has panic disorder. Once it has been established that clinicians should aim to reduce anxiety sensitivity, what are the procedures that will achieve this aim? Here, we must turn to work on active ingredients. By “active ingredients,” we mean the specific interventions that should be included in the psychological treatment in order to maximize outcomes. For panic disorder, an illustrative example of active ingredient research comes from Pompoli et al. (2018), who conducted a component network meta-analysis of published research. The specific procedure of interoceptive exposure was associated with better treatment efficacy and acceptability, whereas muscle relaxation was associated with significantly lower efficacy. Interoceptive exposure has been demonstrated to reduce anxiety sensitivity (Boswell et al., 2013) and to reduce defensive physiological responding to arousal cues (Holtz et al., 2019). Thus, this procedure has been identified as an active ingredient which affects a specific mechanism, resulting in symptom improvement. Parenthetically, we note that the identification of an active ingredient does not necessarily mean that the clinician must provide that ingredient– in panic disorder, for example, there is evidence that cognitive therapy without interoceptive exposure can also be efficacious (Arntz, 2002; Fitzgerald et al., 2021)-- however, we would argue that attending to identified mechanisms of action optimizes efficiency as has been evident in the reduction of standardized interventions for panic disorder from 12 or more sessions to just 5, a strategy that can reduce cost and time barriers to effective treatment (Otto et al., 2012).

Lorenzo-Luaces points to a “long road” from identifying active ingredients to improving clinical outcomes. We suggest that the road need not be particularly long. Consider, for example, Jacobson et al. (1996)’s dismantling study of cognitive therapy for depression, which found that the behavioral activation component was just as efficacious as the more complex package of cognitive therapy. Five years after that trial, Martell et al. (2001) published a manual for behaviorally-focused treatment of depression, an approach that was subsequently confirmed via controlled research showing an improvement in efficacy over cognitive therapy as well as antidepressant medications (Dimidjian et al., 2006). Other examples of successful identification of active ingredients includes the early work of Foa et al. (1980), who found that both exposure and response prevention were needed in order to obtain a clinical effect in patients with obsessive-compulsive disorder. We would argue that such research had, and has, immediate clinical applicability.

One might well wonder (as does Lorenzo-Luaces), given the work that has been done thus far in identifying active treatment ingredients such as behavioral activation for depression or exposure for anxiety-related disorders, why there has not been greater uptake of these ingredients in front-line mental health practice. Lilienfeld et al. (2013) identify several barriers to uptake, including naïve realism (erroneously attributing patient success to a treatment rather than to other factors), misconceptions regarding human nature (e.g., erroneous ideas about the causal primacy of early experiences), statistical misunderstandings (difficulty translating group probabilities to individual cases), placing the burden of proof on skeptics (rather than on the proponents of untested therapies), mischaracterizations of what evidence-based practice entails (e.g., that it is a “cookie cutter” than cannot be tailored to the individual), and pragmatic, educational, and attitudinal obstacles (e.g., discomfort with the increasingly technical scientific literature). Clearly, as noted by Lilienfeld et al. and others, increased education and outreach is needed at both the student and the professional level.

Finally, as Lorenzo-Luaces himself notes, it is most likely a false dichotomy when we pit two forms of research against each other (when arguing that there are research questions that are closer to improving outcomes). Clearly, dissemination and implementation research is critical, e.g., Beidas et al. (2013). However, before such research can be successful, we need to know what to disseminate and what to implement. We know, for example, that cognitive-behavioral therapy (CBT) is efficacious for anxiety-related disorders, and agree that this treatment should be made available to the many people who need it. However, we are also mindful of the fact that only half of the patients who receive CBT for their anxiety-related disorder will actually show demonstrable benefit (Loerinc et al., 2015). There is clearly a balance to be struck here: we need to disseminate what works, and we need to make what works, work better.

We concur with Lorenzo-Luaces that our research needs to take patient diversity into account, and that personalized medicine is the likely future of behavioral health. Ultimately, however, such personalization, including modifying treatment in consideration of patient-diversity factors, requires empirical investigation. Without knowing what works for whom, clinicians and researchers alike are left without direction and may tread the path of tradition or opinion, rather than advancing the efficiency and efficacy of treatments. It is in this spirit that we offer the following thoughts about the importance of strengthening our research efforts to facilitate understanding of why our treatments work (mechanisms) and which aspects of treatment should be emphasized or de-emphasized (active ingredients that engage those mechanisms).

Why is it important to understand mechanisms of treatment?

The assumption at which some readers could arrive from Lorenzo-Luaces’s paper (though Lorenzo-Luaces does not state this, and even at times makes statements that contradict this assumption) is that that mechanism research has been tried and failed. We argue that this assumption does not fit the current evidence; rather, mechanism research has been a rarity in the field and needs greater attention, not less. For example, Edmondson et al. (2018), in a systematic review of NIH-funded behavioral intervention trials to improve medication adherence in adults, found that only 3% of these trials conducted, or explicitly planned to conduct, tests of hypothesized mechanisms of behavior change. Similar limitations have been observed for mechanistic research on treatments for PTSD (Zalta, 2015) and substance use disorders (Magill et al., 2020).

An improved understanding of treatment mechanisms is critical to continued progress in the field. Much of the extant treatment development literature is based on symptoms of a disorder, and we argue that this approach is inherently limited. Knowing that a patient has the symptom of panic attacks, for example, might lead one to develop a treatment focused solely on those symptoms: for example, relaxation exercises to reduce panic. And such an approach may yield at least some benefit. However, understanding the mechanisms behind those panic attacks– for example, that the panic attacks arise due to anxiety sensitivity and catastrophic misinterpretation of physiological arousal– gives us a fundamentally different target and may lead us in an entirely different (and more productive) direction, such as interoceptive exposure and cognitive restructuring around physiological sensations and their meaning. Even with limited available research resources, an exclusive focus on symptoms without attention to their underlying mechanisms may have treatment developers and clinicians “barking up the wrong tree.”

Indeed, Lorenzo-Luaces’s call to abandon aspects of mechanistic science come when major forces in the field are instead moving to intensify a focus on mechanisms. For example, the experimental medicine emphasis evident across many NIH programs was designed to emphasize the importance of mechanism in aiding the progression of clinical science to effective clinical intervention (Nielsen et al., 2018). The establishment of a common fund for the Science of Behavior Change (SOBC) and the continued funding support, via an NIH-wide consortium, of the SOBC Resource and Coordinating Center (https://reporter.nih.gov/project-details/10046157; NIH RePORTER, 2023) has continued to work to enhance mechanism-focused behavior change research by expanding and improving a Measures Repository to encourage standardized measurement and a shared vocabulary for this research (https://measures.scienceofbehaviorchange.org) and continues to collaborate nationally and internationally to promote further advances in behavior change research tools (Stoeckel et al., 2023). SOBC projects also include efforts to promote the standardized reporting of mechanistic research to enhance effective utilization of this research (Birk et al., 2023).

Another ongoing effort is to develop a formal methodology for enhancing the link between therapeutic interventions and their hypothesized mechanisms of action. This includes efforts to refine the taxonomy of behavior change techniques in order to enhance accurate replication of interventions. It also includes specifying intervention content will facilitate faithful implementation of interventions found to be effective, and providing a reliable method for extracting and synthesizing information about intervention content to aid systematic reviews and conclusions about the efficacy of intervention elements. Early outcomes on this process can be found in works by Michie and colleagues (Carey et al., 2019; Michie et al., 2021; Michie et al., 2013), with a variety of researchers providing a longer-term vision of constructing a “periodic table” of mechanisms that underlie behavior change (Rothman & Sheeran, 2020). One goal of this approach is to reduce the methodological flaws and inconsistent methodologies between studies that can hamper firm conclusions about mechanisms in empirical reviews (cf. Morgenstern & Longabaugh, 2000). As reflected in other papers in this special issue, such efforts dedicated to understanding mechanisms is central to intervention development research (Cohen et al., 2023).

Why is it important to understand active ingredients of treatment?

A cursory glance at the list of empirically supported treatments (ESTs) published by the Society of Clinical Psychology (www.psychologicaltreatments.org) will show that a wide range of treatments have garnered adequate research support. However, many early treatment outcome studies did not provide adequate information about what specific interventions were being used, often using only vague terms such as “psychodynamic therapy” or “cognitive-behavioral therapy.” It would be difficult to draw valid conclusions about the effective ingredients of such treatments. The list of ESTs includes more recent multi-component treatments that contain many different interventions within one treatment “package.” CBT for fibromyalgia, as one example of a treatment currently identified as an EST, is described as including education, relaxation, graded behavioral activation, pleasant activity scheduling, sleep hygiene, stress management, goal setting, structured problem solving, reframing, and communication skills (e.g., Glombiewski et al., 2010). Whereas the assessment of such treatment packages is a necessary step in identifying what works, such research does not allow for a determination of which aspects of the treatment are responsible for change (Borkovec & Castonguay, 1998; Gonzales & Chambers, 2002; Henry, 1998). That is, within a given treatment package, there is no way to determine which components of that treatment are therapeutically active or inert.

As a result, practitioners are unable to make informed decisions about which treatment components to use (Herbert, 2003; Rosen & Davison, 2003; Westen et al., 2004), and many treatments may be weakened by ineffective components and/or work for reasons other than those proposed by the treatment developers (Lohr et al., 1998). Rosen and Davison (2003) humorously illustrated this problem with a fictitious “Purple Hat Therapy,” in which a researcher uses an existing EST while wearing a purple hat and declares the existence of a new and empirically supported treatment. Though cognitive-behavioral therapies dominate the current list of ESTs, some have pointed out that there has been little attention to whether the purported mechanisms of CBT are, in fact, responsible for therapeutic efficacy (Henry, 1998; Levant, 2004; Norcross, 1999). We suggest, therefore, that it is time to de-emphasize treatment “packages” that may contain both effective and inert (or even harmful) ingredients, and focus on dismantling and other research that isolates the specific clinical activities that lead to therapeutic change.

Eye movement desensitization and reprocessing (EMDR) is one example of a treatment package whose active ingredients are hotly disputed. RCTs have shown EMDR for PTSD to be superior to no treatment/wait list as well as nonspecific treatment (Rothbaum et al., 2005; Taylor et al., 2003). The most noteworthy element of EMDR is repeated, back and forth eye movements, or alternative bilateral sensory stimulation. However, EMDR also includes elements such as imaginal exposure and cognitive restructuring, well-established elements of CBT for PTSD. Repeated dismantling studies of EMDR have demonstrated that the treatment works equally well with or without eye movements or other bilateral stimulation (e.g., Sack et al., 2016) although this fact has been disputed (see Lee & Cuijpers, 2013). Whereas a thorough examination of this issue is beyond the scope of this commentary, we note simply that if the efficacy of EMDR cannot be attributed to eye movements or similar aspects, and are instead due to aspects of the treatment that are already present in other established therapies, then the addition of this particular treatment package to a list of ESTs would decrease, rather than increase, our understanding of what actually works.

The search for active ingredients is often more complicated than simply determining that an aspect of treatment is inert. In many cases, the search must consider the therapeutic context in which the treatment is delivered. For example, some interventions may be effective when used alone, but are ineffective (or even counter-effective) when used in combination with other interventions. Relaxation and breathing retraining for panic disorder is a prime example. Used alone, these interventions appear reasonably effective for the treatment of panic disorder (Meuret et al., 2003). However, when combined with interoceptive exposure, some data suggest that breathing retraining actually detracts from the treatment’s efficacy (Schmidt et al., 2000), possibly by undermining changes in catastrophic beliefs about physiological sensations. Here, the search for active ingredients can improve the efficacy and efficiency of treatment, guiding a clinician who is using interoceptive exposure to avoid the use of breathing retraining or other relaxation strategies. We propose that improved treatment efficiency is highly desirable: a more efficient treatment makes better use of limited client and system resources, allows clinicians to maximize the number of patients served, and may reduce risk of premature treatment dropout (see also Cougle, 2012).

Conclusions

Although Lorenzo-Luaces, 2023 expresses understandable frustration with the pace and products of research investigating mechanisms of treatment response and active ingredients of treatment, and his points are well taken, we would argue that we need more such research, not less. In his call for greater dissemination and implementation research (which he argues is closer to improving outcomes), he does show us how patients can benefit when we step up our efforts in that domain, but sidesteps the fact that dissemination of an ineffective treatment is no public service. We note that although many treatment packages have been shown to be effective, some of them may contain ingredients that are either unhelpful (e.g., Lohr et al., 1998) or outright harmful (e.g., Lilienfeld, 2007) and more research is needed in order to ensure that our treatments are effective, not only at the “package” level but also at the specific component level. Proper identification of “what works” is needed in order to deliver effective and equitable care to the public.