Personality disorders (PDs) are prevalent disorders (1, 2), with a reported occurrence of 50% in psychiatric outpatient settings (3). Several studies have shown that psychotherapy is an effective treatment for patients with PDs (4–6). However, this patient group is known to evoke particularly challenging emotional responses in therapists (7–12), which makes the treatment more demanding and at greater risk of failing (13). The presence of PD has been shown to have a negative effect on treatment outcome (3) and early treatment termination is more frequent in this group (14). Factors predicting dropout have been found to include commitment to change, impulsivity, and the therapeutic relationship (15, 16). Two recent studies on therapists’ emotional responses (countertransference) when working with patients with PDs indicated that it is not the severity of the patient’s symptoms that appears to be most decisive in their relationship with the therapist. Instead, the patients’ relational problems and personality pathology had a significant impact on the therapist’s emotional response (7, 17). Patients with PD have deep-rooted interpersonal difficulties and tend to unconsciously project more intense and difficult feelings onto their therapists. This can significantly impact the therapeutic alliance. Focus on the relational dynamics is especially important in the treatment of patients with personality disorder (13). However, systematic research on these aspects is insufficient (3).

Therapist emotional response/countertransference (CT) is extensively described in clinical and theoretical literature since Sigmund Freud introduced the concept in 1910 (18). The prevailing view is that countertransference responses involve a joint creation with contributions from both patient and therapist, even though there is an ongoing discussion about the relative contribution from both participants. Although the concept originally derives from psychoanalytic theory, it is now considered an important aspect of the therapeutic process in diverse forms of therapy (9, 19). There are several definitions of CT (20), and it varies whether CT is measured by self-report or observer-ratings. In the present study, we assessed the therapists’ self-reported affective CT response to the patient. This is considered one part of the total CT construct, i.e., the feelings which therapists become aware of, acknowledge, remember, and are willing to report after a session (21).

Empirical research on CT and PD has been limited, but this is changing with increasing number of studies in recent years. To summarize, a consistent finding is that specific PD patterns/diagnoses tend to elicit specific countertransference responses across psychotherapists, suggesting that therapists overall reactions toward patients may be a source of valuable diagnostic information (8, 22, 23). For example, narcissistic PD is associated with angry/criticized and helpless/disengaged CT (24, 25) and borderline PD is e.g. associated with special/idealized (8, 26) and inadequate CT (7, 11, 23) Another consistent finding is that patients with more severe personality pathology is associated with more negative CT (7, 27) and patients at the DSM-IV-TR (APA, 2000) cluster A and B is associated with more negative and varied CT responses than those with cluster C (8, 28).

One of the most consistent findings in psychotherapy research is that the therapeutic alliance (TA; the mutual collaboration and partnership) between therapist and patient is a key predictor of treatment outcomes (29–32). Both patient and therapist-rated alliance have been found predictive (32–37). The capacity to form an alliance is assumed to be a quality that both the patient (38) and the therapist (39) brings to treatment (29). Research suggests that the therapist play a significant role (40), but patients undoubtedly also contribute to the therapeutic collaboration that arises between them (29). Emotional processes are found to play an important role in his respect (41, 42). Several studies have emphasized the importance of the therapists’ ability to recognize and manage their emotional responses for the development of a productive alignment between patients and therapists (therapeutic alliance) - a process where treatment can be completed (20, 43).

Due to their substantial difficulties in interpersonal relationships, patients with PDs face particularly significant challenges in forming a therapeutic alliance (44, 45). Studies have shown that therapists find it especially challenging to maintain a good alliance with cluster B patients (25, 29, 45). A recent study indicated weaker early alliance among patients with borderline PD (BPD) (46). While a strong positive therapeutic alliance is predictive of more successful treatment outcomes, strains and ruptures (episodes of tension or breakdown, that is, countertransference and transference enactments (47)) in the alliance can result in premature termination of treatment (44). According to Hayes et al. (20), acting out of CT is typically harmful, though not necessarily irreparable. Several studies support the view that ruptures in the alliance are unavoidable, and when negotiated, they can present opportunities for therapeutic change (43, 48). However, when the therapist is under great pressure in the moment, the anxiety level increases and can evoke strong feelings of inadequacy and negativity in the therapist. Studies have shown that therapists often react by avoiding these feelings and resorting to maladaptive strategies to manage them (49, 50).

Despite the importance of the relational dynamics (including both TA and CT) in treating patients with PD, few studies have investigated this association. These vary in respect to the operationalization of TA and whether TA is patient-rated and therapist-rated. Most of the studies are correlational in nature and all studies have found meaningful correlations between alliance and different aspects of CT responses (21, 26, 39, 51–55). Only two of these examined alliance and CT in a PD population specifically. One of these, based on the same participants as the current study sample, found that patient rated alliance showed a negative correlation with Inadequate CT and positive correlation with Confident CT (26). Tanzilli et al. (25), investigating therapists emotional response and therapeutic alliance when treating adolescent patients with narcissistic personality subtypes found that lower quality of therapeutic alliance was associated with the grandiose narcissistic subtype (strongly related to the personality syndrome defined in DSM-IV) which was also correlated to angry/criticized and disengaged/hopeless response. Two recent studies have also examined CT and alliance in more detail during the psychotherapy process (43, 56), however, not in a PD population. In their sample of depressed patients, Falkenström and Holmqvist (56) found poorer working alliance and less engagement among patients where the therapist experienced negative feelings in sessions. Therapists’ positive feelings related to a better working alliance, patients feeling more engaged, and therapists’ positive feelings also had a positive effect on outcome. Tishby and Wiseman (43), in their sample of students with mostly mild to moderate depression, and difficulties in relationships, found that therapist negative CT patterns were associated with more alliance ruptures and less rupture resolution with patients. The results indicate that therapist’s own unresolved conflicts may impact countertransference negatively, and the authors suggest that increased countertransference awareness helps therapists recognize and address ruptures, initiating resolution processes (43, 57).

In meta-analyses, early alliance has been classified as first to fifth session, except from Frank and Gunderson’s long-term study (58, 59). In their study, alliance measures taken in the 6^th month of treatment were considered “early”. The present study also employs a 6-month measure, as many of the therapies in the present study are long-term, lasting up to 60 months at the most. As far as we know, no previous study has investigated patient-rated early alliance as a predictor of therapist CT longitudinally in therapies with PD patients.

Early treatment termination is more frequent among PD patients and represents a prevalent issue in clinical practice. The reasons for non-completion in psychotherapy are multifaceted and include situational, personal and relational factors, including the therapeutic relationship (50). More studies are needed to investigate what happens in the interactions with PD patients who do not complete therapy. Particularly relevant is therapists’ countertransference (CT) during challenging therapies. In the present study, completion of psychotherapy according to scheduled plan was the chosen outcome. To the best of our knowledge, only one study has examined CT and treatment course in a sample of patients with PD specifically. Røssberg and co-workers (28) found strong correlations between CT feelings and clinical change during treatment. Patients who dropped out of treatment evoked significantly more negative CT reactions in the initial treatment phase than patients who completed treatment.

The present study explores the essential therapeutic dyad in treatment of patients with considerable emotional, relational and attachment issues. As a naturalistic, clinical study, the sample is large, and representative of PD-focused treatments as applied within specialist mental health services in Norway.

The main findings are the following: Firstly, the demonstration of significant, independent associations between early negative CTs and lower early patient rated alliance. More specifically, lower ratings of WAI were associated with lower therapist rated Confident and Idealized CT and higher initial Inadequate CT, and conversely, higher early WAI with significantly higher initial Confident and Idealized and less initial Inadequate CT levels. It thus seems that the quality of early WAI as experienced by patients was largely in concordance with the quality of therapists’ experiences in the same phase of treatment, which is in line with other correlational studies on CT and alliance (21, 26). Moreover, the study demonstrated that the relation between early alliance level and CT response levels remained stable throughout therapy. That is, high or low levels of early alliance did not predict any deviating pattern of CT development during therapy. If alliance was low at the start of treatment, our models indicated that positive CT responses would start from a lower level, follow the general change trend in the sample and increase over time, but this lower level would still prevail throughout treatment. Our findings underline the importance of the quality of the starting point in therapy. Investigation of parallel developments of alliance and CT during the therapy was outside the scope of the present study but could further illuminate qualities of the change process in PD treatment.

It should be mentioned that overall, patient-measured alliance was high. Although early alliance had a highly significant impact on CT, the actual change in intensity on the CT scale was low. Possible explanations for the predictor-associated small nuances in CT intensity are the same as those described in a previous publication (7). That is, relatively infrequent measurement points (e.g. no immediate reactions after sessions) very likely contribute to less intense emotional responses on the CT scale. Nevertheless, the levels are comparable to those reported in other studies that have examined CT, using various versions of FWC as instrument. The significant findings in the present study are interesting because they show two non-complementary perspectives, namely the patient and the therapist, and demonstrate the importance of CT and the interaction between the two participants in the therapeutic dyad.

The relatively high early alliance ratings might seem surprising. Patients with PDs are known to struggle with significant impairment in interpersonal relationships and there are often problems in the formation of a therapeutic alliance (44, 46). However, several other studies examining patients with severe relational problems have reported alliance (WAI) within the same range (74, 75). In a study by Folmo and colleagues, BPD and alliance development were examined. The patients in their study are from the same quality register as the present study. They found WAI levels within the same range and the overall picture of change over time was a significant increase of all three working alliance subscales (75). There might be several reasons for the relatively high WAI scores in general. It might illustrate the potential value of PD treatment teams with experienced therapists in specialist health services. It is also natural to consider other explanations. Another study, using a different alliance measure (CALPAS) found that therapists’ alliance ratings were consistently lower than patients’ rating over time (45). This might also be the case in the present study; however, we do not have therapists’ assessments of alliance. Lastly, the phenomenon of self-serving bias is always present when using self-report as a method. That is, it is possible that the patient may not be able to be honest and may find it difficult to report negative feelings about the therapist. Research suggests that self-report instruments may be particularly problematic for patients with PDs (especially, narcissistic, paranoid, and antisocial), who often lack self-awareness and tend to defensively downplay or deny their psychopathology (25, 76). Our study has, however, not differentiated between different types of PDs in relation to alliance nor examined the longitudinal alliance development. To summarize, even though the overall alliances scores were generally high, lower early alliance was significantly related to more negative CT in therapists. This conveys the idea that the impact of early alliance continues or remains relevant for the therapeutic relationship during the course of therapy. As to our knowledge, no other study have investigated this association.

The second main finding is the demonstration of significant associations between the developments of increasingly negative CT responses in therapists during treatment with non-completers as compared to completers. The negative CT development in non-completed treatments was irrespective of the quality of early alliance. In the previous study on CT from the same quality register (7) we found that, overall, patients who did not improve were associated with more negative CT. The previous study did not investigate treatment completion as an aspect of outcome. Nonetheless, meeting to therapy and completing according to the scheduled plan, is likely related to the treatment process itself. Several studies suggest that patients with more PD pathology elicit more negative feelings in therapists (7, 12, 21, 22) and clinical studies report high drop-out rates for PD patients in general (77, 78) and BPD in particular (79). Our study has not delved into further investigations of mediating or moderating variables of non-completed therapies, such as the type of PD, severity of PD, or other comorbid conditions as mentioned above. This would be beyond the scope of this study, but will be important aspects to investigate in future studies. In contrast to Røssberg and coworkers study (28), which found that dropouts very early in therapy were associated with more negative CT, the present study shows a more consistent early baseline regarding CT levels. Differences emerged more over time, with increasing negative CT in the non-completer group. This may indicate that something happens in the process along the way, which is not necessarily captured at the start of therapy. It should be noted that the present study and Røssberg and coworkers study are different in several ways, and thus not directly comparable, e.g. “early” in our study refer to WAI at 6 months. Further, the study designs are different, with correlational design versus longitudinal analyses in the present study. In addition, their study examined drop-outs, while the present study examines a somewhat wider definition of non-completer (dropped out or ended therapy of different reasons).

The therapists in this study were generally experienced, working in specialized units, where CT is part of their daily routine and systematic supervision is recommended. Despite their experience, the study shows they are negatively affected over time by patients who do not complete treatment as planned. However, we do not know whether negative CT predicts non-completion. This is an interesting question that should be addressed in futures studies. Further studies of the psychotherapy process are needed, including mutual alliance, and effects of therapist’s supervision and training on negative CT. However, this quality register with few assessments over time did not allow for such detailed investigations. It could be mentioned that there is an increasing research interest in examining treatment non-completers and the impact of the therapeutic relationship, with ongoing research in this area (50).

Comparing the non-completer group to completers, the former was generally characterized by significantly greater PD comorbidity and overrepresentation of paranoid PD. This may not be surprising, as the specific features of paranoid PD are likely to be of significance for the formation of a trusting therapeutic relationship, thus potentially complicating the process of therapy. Comorbid paranoid PD features, are, however, quite common in more severe PD conditions presenting to health services. The non-completer group was also characterized by more individuals with PTSD, OCD, eating disorders and substance disorder, overall indicating that the non-completer group represented a poorer functioning patient group. This illustrates the important clinical challenge of such severe conditions. A recent study of severely self-harming PD patients in repetitive, inpatient situations, were indeed characterized by such complex comorbidity (80). Interestingly, as not correspondingly reflected in therapist ratings of CT, the non-completer group and the completer group differed on patient-reported alliance with significantly lower early alliance ratings in the non-completer group. However, as mentioned, the mean scores were relatively high, and score levels in both groups were within a satisfactory range. Thus, inferences from WAI scores alone may not be sensitive enough in identifying patients with risk of non-completion. The discrepancies between patient-rated alliance and therapist-rated CT underline that a more fine-tuned investigation of the starting point in therapy, patients’ vulnerability and therapists’ competence could reveal more nuances on how the process develops and the mutual understanding and bond in therapeutic dyad develops.

Patients with PDs still pose significant challenges to psychotherapists. Their repetitive interpersonal struggles are often actualized in the therapeutic relationship, hindering the formation of a positive alliance and evoking difficult emotional responses in therapists, which puts therapists’ performance under pressure. There is a growing body of research on strategies to enhance therapists’ abilities to navigate these pervasive challenges. In general, therapists are advised to engage in reflective practices that increase their self-awareness, affect regulation and interpersonal sensitivity (43, 57, 81). Meta-analytic evidence (20) suggest that therapists may prevent countertransference enactment and reduce non-completion if they attend regular supervision, develop emotional regulation strategies, such as mindfulness-based techniques, set healthy boundaries with their patients, adopt non-defensive stances and practice self-care, to name a few. Furthermore, therapists are encouraged to commit to ongoing personal therapy to address own, unresolved issues. Numerous training programs are currently being developed to enhance therapists’ abilities to tolerate intense emotions, including negative ones such as anxiety, anger, and stress e.g (82, 83).

In conclusion, the present study shed light on the underlying CT processes in PD treatments, emphasizing the important impact of relational factors in the bi-directional nature of the therapeutic relationship. Finally, it must be emphasized, that even though associations between CT, alliance, and treatment completion have been identified, we cannot make any claims about causality. The study highlights the need for further, longitudinal studies on the combined influence of therapist and patient factor in the therapeutic relationship. This may contribute to the development of more tailored interventions for individuals with personality disorders (50).

The study supports research indicating associations between a patient’s perception of alliance and therapist CT responses. Our study underlines the importance of the quality of the starting point in therapy. In addition, the study showed that non-completers were associated with increasingly negative therapist CT during treatment. This result was independent of variation in early alliance. This underscores the significance of continuous countertransference awareness when working with personality disorder patients. Further exploration revealed that comorbid paranoid PD, PTSD, OCD, and eating disorders were more frequent in the non-completer group. The study contributes to psychotherapy research emphasizing the value of the emotional and relational quality of psychotherapeutic relationships in the treatment of patients with PDs.