The Elephant on the Couch: Side-Effects of Psychotherapy

Although we take it as a given that many psychotherapies are efficacious, form a cornerstone of much current practice, and are valued by many patients, there is a dissonance in the way in which physical therapies and psychotherapies are considered in terms of their cost–benefit ratio. Any potent intervention has both the capacity to cure and to harm. For drug-based therapeutic trials, adverse event monitoring is mandatory. By contrast, evaluation of psychotherapy has historically weighted the ‘benefit’ side of the equation. For example, is a particular type of psychotherapy effective, is one psychotherapy superior to another, or does psychotherapy benefit a particular condition?

At first pass it might appear bizarre to question whether psychotherapy could be harmful or have substantive side-effects, with Nutt and Sharpe recently observing that there is an ‘assumption … that as psychotherapy is only talking … no possible harm could ensue’ [1]. Certainly, patients rarely raise such concerns. By contrast, when a psychotropic drug is prescribed, most patients inquire about likely drug side-effects, while medico-legal injunctions oblige the practitioner to detail and document substantive side-effects.

Why the dissonance across those treatment modalities if we accept the principle that all effective treatments risk adverse events? In arguing against the common assumption that adverse events of psychotherapy are slight, we offer several examples to argue that substantive costs can emerge from both acts of omission and commission.

Impact of inappropriate psychotherapy

If psychotherapy is provided as the only or principal therapy for a condition for which it is either inappropriate or ineffective, the patient may be exposed to a lengthy period of ongoing symptoms and disability – an adverse outcome. As ‘acts of omission’, such paradigm failures can be distinctive, with the Osheroff case being a well-documented example.

As detailed by Shorter, Osheroff, a 42-year-old physician, was admitted to Chestnut Lodge with symptoms of psychotic depression, received near-daily intensive psychotherapy and, over his 7 month admission, was denied medication despite his own requests [2]. Subsequently transferred to another hospital, he recovered after receiving psychotropic medication, although his wife had left him, he had lost his hospital accreditation and his medical partner ‘ousted him from their joint practice’ during his extended hospitalization. Osheroff sued for malpractice on grounds that he should have received medications of demonstrated efficacy rather than intensive psychotherapy.

Impact of inappropriate psychotherapist behaviour

Psychiatric patients are commonly highly troubled and vulnerable, whether personality based, and/or a consequence of their illness. An insensitive, critical or sexually exploitative therapist may increase a poor outcome risk. The experience of Anna O, the first patient of the cathartic method of psychoanalysis and dynamic psychiatry, and who was exposed to inappropriate psychotherapy, has been detailed extensively. Weissberg stated that ‘Breuer's interventions make it possible that he unwittingly encouraged and amplified Anna's dissociations, reified her ego fragments, and then explained Anna's symptoms with the pseudo-memories and confabulations recovered from Anna while she was hypnotized’ [3].

While Rebekah Beddoe's book titled Dying for a Cure is promoted as demonstrating the inappropriate use of psychotropic medication, the perturbing psychotherapy (provided by ‘Max’) is at least equally concerning [4]. A few quotations, recounted by Rebekah without apparent irony, capture the ‘creep’ of an inappropriate and exploitative psychotherapist:

According to him I deserved a lot more love and attention than I had ever received. Letting go of hurt and guilt while Max held me and absorbed my sobs was narcotic … ‘Take off your shoes and socks and pop your feet up here’. He patted his lap. ‘Now tell me this doesn't feel delicious’. He gently circled each toe round and round, and stroked the sole of my foot. It felt delicious alright … Right out of the blue one day, he wanted me to describe an orgasm … Max came to my rescue and helped me along … He introduced an exercise that I termed ‘cuddle therapy’ … He gently pressed his hands into the small of my back and pulled me in close.'There, it's only a hug – I'm not trying to fuck you’ … Each session was to start with a greeting just like this. I'd be critiqued on my ability to make body contact. Further on she recounts his enquiry: ‘Tell me, Bek, what does your clitoris look like?’.

Prevalence of adverse outcomes in psychotherapy

Bergin in 1967 coined the term ‘deterioration effect’ to describe how ‘psychotherapy may cause people to become better or worse adjusted than comparable people who do not receive such treatment’ [5]. Foa and Emmelkamp focused on treatment failures, examining factors such as refusal, dropout, non-response and relapse; diagnostic and assessment error, inadequate application of a treatment programme, the patient's personality, motivation and difficulties in the therapeutic relationship [6]. Deterioration in psychotherapy has also been viewed as ‘not only worsening symptoms, but lack of significant improvement when it is expected and even the acceleration of ongoing deterioration’. The idea that psychotherapy could be deleterious has historically been met with both inertia and opposition [7].

Quantitative studies are few, limited in scope and weighted to idiosyncratic psychotherapies or to their more problematic or peripheral application. A few empirical studies have quantified the broad proposition. For example, it has been estimated that approximately 3–10% of patients become worse after psychotherapy, with slightly higher rates (7–15%) quantified for patients with substance abuse [8–10]. A recent article suggested that approximately 10% of individuals worsened after commencing psychotherapy [11]. It is clearly difficult, however, to establish the percentage of those who would have worsened regardless of psychotherapy. Additionally, few studies go beyond documenting deterioration in primary outcomes, to consider alternate adverse outcomes such as new symptoms, increases in anger or negative family effects.

In the next section of this paper we report representative studies from the sparse literature base.

Harmful effects reported for specific psychotherapeutic interventions

Werch and Owen reviewed preventive interventions for substance use in youth and young adults, and found 17 studies with documented negative effects (e.g. increased substance use and a reduction in self-efficacy) [12]. Negative effects have been documented when interventions use ‘resistance skills training’ without normative education, with the Drug Abuse Resistance Education (DARE) prevention programme judged as increasing substance use [12]. In ‘deviancy training’, where deviant behaviour is modelled and reinforced within a peer group setting, negative ‘iatrogenic’ outcomes have been described [13, 14].

It has been suggested that therapists who induce high emotional arousal may inadvertently cause an increase in alcohol consumption, especially in those with comorbid mood disorders [14]. Interventions that risk increasing a person's feeling of being stigmatized or in which they are blamed for not meeting intervention targets, have been held to increase helplessness and self-blame, and so undermine self-efficacy [15]. Moos suggested that clinicians need to be cautious with substance-abusing patients when using high-risk treatment processes such as confrontation, criticism and highly emotive techniques, because they can exacerbate primary symptoms, or initiate new symptoms such as increased anxiety or anger [9].

Szapocznik and Prado detailed how interventions may have adverse effects on families and friends [16], particularly if the individual undergoing therapy becomes more self-absorbed or self-centred [17–19]. In a small randomized controlled trial of a psychoeducation group for partners of those with bipolar disorder, it was quantified that, although partners improved their knowledge of the illness, the anxiety levels in the ill partners increased [20].

Lilienfeld has provided examples of ‘probably harmful’ and ‘possibly harmful’ psychological treatments [21]. The first group included the ‘Scared Straight’ programme, which exposes at-risk adolescents to the realities of prison, and in which it was established that those receiving such interventions were significantly more likely to offend [22]. Another example was critical incident stress debriefing (CISD) targeted at post-traumatic stress disorder (PTSD) and anxiety symptoms in people exposed to severe stressors. One randomized controlled trial of burn victims quantified an increase in anxiety and PTSD scores in those assigned to the CISD intervention compared to those in the control group [23]. Similarly, in a 3 year follow-up randomized controlled study of motor vehicle accident victims who received CISD, treated subjects exhibited higher travel anxiety and global pathology [24]. In a review, Bledsoe highlighted the risk of worsening stress-related symptoms in both patients and personnel, and concluded that it should never be a mandatory intervention [25]. The same risk has been attributed to grief therapy. In a meta-analysis of 23 randomized controlled trials, Neimeyer reported that 38% of patients undergoing grief therapy may have done well if they had not received treatment [26], although this analysis is controversial [27]. In the ‘possibly harmful’ treatment group, Lilienfeld included group interventions for antisocial behaviour based on deviancy training, as noted in the previous section [14]; boot camp programmes for adolescent and adult offenders [28]; as well as debriefing and rebirthing strategies.

Substantial controversy surrounds psychotherapy for false or repressed memories. Reviews have generally failed to provide evidence that traumatic memories are any more likely than non-traumatic memories to be repressed – or that they can be reclaimed via techniques such as guided imagery and other suggestive therapeutic procedures [29]. The consequences of false memory therapies have engendered heated debate. Although there are no controlled studies substantiating destructive effects of memory recovery techniques, legal claims refer to increases in suicidality and psychiatric hospitalizations [21].

Dissociative identity disorder therapy uses suggestive techniques to reveal and promote interaction with other latent identities or ‘alters’. Recovering memories of childhood sexual abuse is sometimes part of the intervention. Similar to concerns about the creation of false memories, there is disquiet about such techniques creating false identities, with associated self-harm and aggressive behaviour increasing symptoms [30].

Although the literature tends to focus on idiosyncratic psychological treatments, we now consider the theoretical potential of certain mainstream psychotherapies to induce adverse events, because there has been little other than anecdotal reporting.

Adverse events that might be non-specific to the type of psychotherapy

Illness status itself provides a fulcrum for psychotherapeutic engagement. Both physical or psychological illness are associated with a set of reactions, which include a sense of disconnection from one's usual world, and a loss of (i) the sense of indestructibility (or omnipotence), (ii) the competence and completeness of one's reasoning, and (iii) control over oneself and one's world [31]. In response, the sufferer might be expected to seek a therapist to trust, lean on to varying degrees (i.e. between normal and pathological dependency) to reduce their sense of isolation, and advance the return of control. ‘Illness’ status may, however, be perpetuated by the patient and/or the psychotherapist, allowing secondary gains to accrue from the ‘sick role’ status, prolonging psychotherapy beyond what is ‘necessary and sufficient’. Such issues link to the concept of ‘dependency’.

The longer any patient attends a psychotherapist – irrespective of how therapeutic the therapy – the patient risks contracting their independent capacity to make decisions (self-mastery), whether by deferring in sessions to their therapist or by filtering decisions outside therapy through the therapist's decision-making model. The risk is for the patient to remain in a therapeutically shaped ‘comfort zone’, distanced from the capacity and risks inherent in making their own mistakes in the real world and, more importantly, learning from them, and so shifting their interpersonal investments to limit primary and extended relationships. In the context of psychoanalytic inpatient treatment of borderline personality disorder, Chiesa et al. described that elements of long-term inpatient treatment ‘might carry the risk of iatrogenic and anti-therapeutic effects for a sub-group of patients’ [32].

The Cambridge–Sommerville study offers some empirical supportive data. The initial study of 650 pre-delinquent boys compared counselling and support from case workers and a control condition [33]. Follow up (up to 17 years) showed a trend for boys having the active treatment to be more likely to have gone to court and to record more offences. Those whose counsellors visited them the most were the most likely to fare badly compared to the control group [34]. Study results are clearly capable of many explanations. It is possible that the need for counsellors to visit the higher risk boys more frequently accounted for the dose-related negative effects of treatment. Alternatively, people may benefit less from being advised how to proceed through life and more from learning as much from their mistakes as from their successes. If the latter explanation is valid, it supports concern about a patient or client becoming dependent on a therapist. For individuals who have a dependent personality style, limited social supports and networks, and/or chronic and disabling conditions, the risk of excessive dependence – and maintenance of a sick role – is clearly higher.

The Scylla–Charybdis dilemma is clear. Some degree of dependence is necessary in the psychotherapeutic alliance to allow the healing common factors to produce their benefits – but its potential to undermine self-mastery is substantive, can occur early and risks increasing over therapy. The issue of dependence and its potential for inducing harmful side-effects is well recognized by experienced practitioners and leaders in the field [35].

Adopting Frank's model [36], we can formulate general reasons as to why common factors might, if not optimal, contribute to adverse events and outcomes. First, there is a need for an emotionally charged confiding relationship and a healing setting. Adverse outcomes might emerge if the therapist is so passive or inert as to prevent activation of such therapeutic ingredients (including hope) or if the therapeutic setting has limitations. Second, if the therapist prioritizes their own needs (e.g. exploitative, narcissistic, control, voyeurism) over the priorities of the patient, the healing setting is compromised. Third, although a treatment logic contributes to a good outcome, many psychotherapists may recommend psychotherapy or therapy without specifying why that modality is salient for the patient and, perhaps more importantly, fail to lay out a road map or set of objectives for the therapy. Fourth, in relation to a credible therapy, although many patients are unlikely to take up or continue with a treatment that lacks credibility, individuals with psychological distress are often so perturbed by their condition that their judgment about such matters can be compromised. If in doubt, they may continue with the therapy, due to the belief that the credibility or benefits of the therapy will emerge over time, or that there's something wrong with them, or because they are unaware of alternative strategies [37].

We now note some theoretical adverse outcome risks to specific psychotherapies.

Psychodynamic psychotherapy

The high session frequency and extended period of much psychodynamic psychotherapy make the issue of dependency particularly salient. Further, there may be no short-term end-point – with unfinished business being the diffuse agenda. For those who intellectually or otherwise enjoy pursuing self-awareness, such psychotherapy may meet multiple other needs, so mitigating adverse consequences. The self-absorption engendered by the process, however, can lead to individuals weighting the intellectual and self-exploratory components above real world issues, thus risking a sterility to their life – as is observed in any individual who narrows their world to pursue a narrow hobby or career track.

The retrospective focus on historical factors (as against dealing with current issues) may promote an externalized locus of control if the person is encouraged to conceptualize their difficulties as arising from a fixed external event or individual. A potential consequence of externalizing attributions of current difficulties to the behaviour of others (particularly parents) is estrangement, disengagement and passive adoption of the victim role.

For some, increasing intellectualization and viewing their therapist as invariably wise and infallible risks ‘intellectual incest’, and a decreased capacity for independent judgment. The therapeutic sessions can take on primary importance and become a self-absorbing safe retreat that replaces active participation in real relationships and narrows the ‘lived life’.

Many analysts practice a reflective style, avoiding responding to any interpersonal nuance. While cogent supportive reasons have long been argued, the stratagem risks being viewed by the patient as lacking empathy, being at variance with the style of communication that underlies usual reciprocal and rewarding human interaction. For many patients such perceived distance can reify their doubts about their interpersonal skills and self-worth.

Some patients describe analytic psychotherapy as providing a ‘heads you lose, tails you lose’ model. If you abandon protective defence mechanisms and declare frailties, you are exposed; if you deny, you demonstrate resistance. Both analytic and cognitive therapies provide cogent explanations for an individual's distress. Individuals with personality disorders who have rigid and extreme schemas frequently struggle to compare their perceptions with those of the expert, and are forced to accept or reject these without the capacity for them to be integrated. The resulting dissonance between inner experience and the imposed perspective can risk bewilderment and further instability [38].

Transference is a common component of analytic psychotherapy and is useful for the analytic psychotherapist to understand the patient and for the patient to acquire insight into nuances of earlier relationships. Theoretical risks include promoting the omnipotence and omniscience of the therapist – and the comparative frailties of the patient. Over time, long-term bonds of attachment to the therapist may make termination of therapy a traumatic life event, particularly if transference has been an important therapeutic component.

Evidence-based psychotherapies

Both cognitive behaviour therapy (CBT) and interpersonal psychotherapy (IPT) have been termed ‘evidence-based’ in that they have been subjected to multiple randomized controlled trials, particularly as treatments for depression, and with support for their efficacy [39, 40]. They consequently have high cachet value and are often positioned as first-line therapies for a range of conditions, including depressive and anxiety disorders.

Because such treatments have an underlying logic (i.e. CBT being designed to modify underlying cognitive schema, IPT focusing on conflicts and transitions in patient relationships and social support network), have a template for proceeding, are commonly manualized, and are generally time-limited treatments, some of the adverse event risks listed earlier would appear less likely: in particular, propagation of the sick role and dependency.

As noted, IPT focuses on social factors. Although distal and proximal antecedent social factors often predispose to and/or precipitate psychiatric conditions, they do not necessarily provide the most salient fulcrum for intervention. If the psychiatric condition is primarily biological (and preferentially responsive to medication) or psychological (reflecting, say, primary personality problems), then IPT may resemble a gardener who waters the flowers but does not consider whether the garden needs fertilizing or what plant might be biologically suited for the actual garden plot.

CBT assumes that the individual has an ongoing cognitive schema that causes them to view themselves, the world and their future with negative ascriptions. Therapy is designed to challenge their cognitive assumptions and encourage behavioural repertoires generating more positive outcomes. The focus on rational thinking assumes a certain level of reasoning capacity – which may be lacking due to low intelligence or current symptoms. Some patients confronted with such expectations – and unable to meet them (particularly as a consequence of severe depression) – may have their sense of self-worth further undermined. Further, CBT shifts responsibility onto the individual for active engagement and conduct of the techniques. A recipient may feel guilty if treatment does not result in the expected improvements, without realizing that there are many other factors that may affect response.

Some experienced cognitive therapists suggest that CBT can be toxic to some individuals, particularly those with obsessive personalities, by increasing worry and introspection, fuelling rather than relieving anxiety and depression. Vulnerability to such adverse events may be a consequence of stage of illness [41, 42]. In bipolar disorder, CBT benefited those individuals in the early stages of illness, while those people who had more than 12 prior episodes of illness actually deteriorated with CBT [43]. This suggests that the progressive neurostructural, cognitive psychological or social factors that change with the course of illness, may alter the pattern of response to and the benefit:risk ratio of CBT.

Therapist style

Independent of the therapeutic modality offered, therapist style is a major influence on outcome and adverse events. Interpersonal characteristics of the therapist as expressed to the patient in therapy may promote or compromise therapeutic alliance. As noted, common factors promoting good outcome are empathy, respect for the patient, confidentiality, a declared logical therapeutic rationale, instilling rational hope and providing a healing setting.

Again as noted earlier, the therapist who is exploitative, overly narcissistic, patronizing, uncaring, inattentive (e.g. asleep during sessions or not remembering key details of the patient's history), or unable to establish some congruence with the patient and their world, may be expected to create a lack of fit and an adverse outcome. According to Horowitz, ‘The rare therapist who is a malignant narcissist is capable of inflicting severe damage by sadistically exploiting the group to satisfy his or her own pathological needs’ [44]. Sexual boundary transgressions are the most overt noxious example.

Particularly in group therapies, a charismatic but confrontational therapist who demands self-disclosure, emotional expression and change in attitudes may be responsible for deterioration in a participant who feels unnecessarily exposed and vulnerable [45]. Conversely, an overly consoling therapist may encourage dependency and helplessness. Hoag et al. suggested that positive effects of group therapy for adolescents may be masked by ‘psychonoxious’ therapist factors, such as very authoritarian and prematurely demanding therapists [46]. Recognition of such issues has led to recommendations that therapists explore countertransference issues that may impact on outcome [35].

In any therapy situation there are personal and relationship factors that may not only affect positive outcome, but also contribute to harmful effects [47].

Why is the adverse side of the ledger neglected?

If these exemplars of omission and commission are accepted as potential cost risks to psychotherapy, why do we neglect this side of the ledger? First, it may be that we assume that the caveat emptor principle holds – that if a patient is referred to a psychotherapist who is clearly ineffective, exploitative or insensitive, they would choose not to return, thus preventing exposure to any distinct adverse event. For those who chose to stay, however, two processes may occur that, because they are neither overt nor clearly causal, may not be appreciated as generating adverse events. First, the ‘boiling frog’ principle, in which we adjust to stressors if they occur incrementally or slowly, and become accepting. Thus, when omission and commission concerns are less evident, blatant or immediate, a patient may continue with the psychotherapy despite a progressive smouldering enmeshment process that, because it unfolds slowly or subtly, builds to the boiling frog analogy. Examples include an unstructured meandering psychotherapy that fails to address the patient's problems, or the therapist subtly prioritizing their own needs. Worse, the patient may be unaware of the exploitation and, as one consequence of the confused agendas, even enjoy it. As Beddoe observed: ‘Within days Max's visits became the most anticipated event in my day’ [4].

A second contribution is that, while there is usually a clear-cut causal process in establishing a drug side-effect, it is less easy to argue any temporal causal link associated with psychotherapy. For example, if a depressed patient is commenced on an antidepressant drug, and they report immediate sedation and weight gain, the drug is the a priori causal agent. For a depressed patient receiving ineffective or inappropriate psychotherapy, negative consequences lack the immediacy of a distinctive drug side-effect. Even if the patient feels some discomfort about the psychotherapeutic approach and/or the psychotherapist themselves, there is a risk that such concerns will be rationalized (e.g. ‘I'm aware that therapy will take a long time’; ‘I'm not so sure about my therapist, but maybe that's my fault’) rather than being linked to something lacking or inappropriate in therapy.

Discussion

We suggest that evidence-supported treatment status requires not only an examination of efficacy but analysis of how well these interventions translate into real world contexts – their transportability – and which should include both their clinical effectiveness and risk of adverse events. Such research is not only of integral importance but also allows a more considered weighing up of the cost–benefits of prescribing a psychotherapy.

In this paper we have proposed that psychotherapy may risk adverse outcomes both as a consequence of the therapist and of the therapy. It could be that the first proposition is unjust, both on theoretical grounds (i.e. akin to an individual criticizing religion on the basis of disliking their local minister) and on an equity basis (i.e. efficacy studies of psychotropic drugs do not examine interpersonal characteristics of the prescriber). But the practice – and much of its benefit – of psychotherapy is dependent on the practitioner prescribing themselves. If prescribed optimally, the patient's propensity to benefit is advanced, while if suboptimal or toxic, then the patient risks an inadequate response or an adverse outcome.

It could be argued that all treatments that risk probable and possible harm should be prioritized for such clarification. Lilienfeld went so far as to suggest that identifying harmful treatments may be even more important than identifying beneficial ones [21]. We would argue, however, for evaluation of adverse events across all psychotherapies. First, this would allow identification of integral risks across all psychotherapies (whether therapy related or therapist related). Second, it would allow identification of therapy-specific risks. For example, are long-term psychotherapies a greater risk for engendering harmful dependency, and are CBT or IPT associated with noxious outcomes in certain definable circumstances?

We argue then for strategies that identify both generic and psychotherapy-specific adverse events to be implemented. Together with efficacy data, such information would provide more precise process and context information about the ecological niche of differing psychotherapies (i.e. what type of therapy risks adverse events in what type of patient).

Such data would best be derived from formal randomized controlled efficacy studies, clinical effectiveness studies and real world clinical practice, while the last would beneficially examine for adverse events both retrospectively (i.e. after therapy has been completed) and longitudinally. Longitudinal evaluation allows that quite differing adverse event risks may be compartmentalized to or overrepresented at differing stages across therapy. Inclusion of the derived measure in treatment studies (whether of psychotherapies, drug or other therapies) would then allow a much richer opportunity to partition the influence of so-called specific and non-specific therapeutic components on outcome.

Any such measure might be optionally incorporated into day-to-day clinical practice. For example, some therapists might appreciate a checklist of potential risks in order to calibrate their interactions with patients over the course of therapy. Quantifying an individual therapist's dropout rate against a standard might inform a clinician as to whether their practice profile was aberrant. Having a subset of patients complete an anonymous structured questionnaire would provide the clinician with information allowing corrective strategies.

This paper takes as a foundation that psychotherapy is an efficacious cornerstone of current practice. The very potency of such therapy gives rise to risks that may not have been adequately appreciated, and thus there has been a tacit assumption by practitioners and patients that psychotherapy is largely devoid of risks. This may be a double blind. We suggest that there is a need for greater awareness and appropriate monitoring of risks, and that pursuit of this proposition will advance the risk–benefit ratio of psychological treatments.